mirror of
https://github.com/garrytan/gstack.git
synced 2026-11-02 23:14:33 +01:00
Garry Tan
55f8151fbe
* docs: approved reliability follow-ups plan (Oct 2026)
* D1: score overlay wrong answers as comparison results (contract v4)
A completed answer that fails the fixture's output check is now a valid
measurement with taskCorrect false and answerError set, in either arm.
Only execution, scope and metric failures invalidate a trial. ON
correctness still gates the case, so an OFF wrong answer stays in the
comparison. OVERLAY_CONTRACT moves to v4; v3 verdicts are not
reinterpreted. Replays the 37198445662 sonnet trials (OFF-0 wrong).
* C4: score auq-matrix substance with a 3-sample judge panel
Each skill keeps one native capture in the same execution; the
recommendation-substance grader now runs as a JUDGE_PANEL_SAMPLES panel on
that capture and gates on the mean against the unchanged minimum of 4. A
failed sample fails the panel without resampling. Approved at the gate
2026-10-04 for censuses after this change.
* E1: one Bun version source; setup refuses below the 1.3.3 floor, updater checks before advancing
bin/gstack-bun-version.sh holds the security floor (1.3.3) and the tested
version (1.4.0, the CI pin and engines.bun). setup refuses Bun below the
floor before writing anything, warns between floor and tested, warns and
continues on an unparseable version, and prints a 1.4.0 install hint.
Prereleases sort below their release (semver).
gstack-session-update now fetches, reads the incoming release's floor and
checks the bun setup will run before fast-forwarding to that exact
revision; below it the update is held with reason bun-too-old and a
resume command, leaving the revision and installs untouched.
Fixture copies of setup (team-mode, claude-code migration) now include the
sourced file; the slow-git lock test sleeps on fetch, the updater's
network step.
* E3: record the owned Chromium's PID over CDP so the #2709 reap runs on Playwright 1.62
Playwright 1.62's Browser has no process(), so chromiumProcInfo was always
null and browse stop never reaped a surviving headless Chromium. The owned
headless launch now reads the browser PID from CDP SystemInfo.getProcessInfo
(bounded at 2s; a hung or failed call logs once and records nothing). The
reaper's start-time and cmdline identity checks are unchanged, so a recycled
PID or an externally owned browser is never signalled.
* A1: ship docs gate records post-child hashes with gstack-docs-candidate, saves each record once
Stored transcripts (37198445662 gate-census-2) show the parent's reads of
release files take ~1 s; the parent's own model time dominates (completion
268 s of 392 s, store ~270 s of 387 s), and 45-78 s of it is retyping hash
maps, child output and records into several artifacts after the child
returns. The child contract is unchanged.
* A2: /plan-ceo-review sends the mode handoff line before any other tool call
plan-ceo-mode-routing's HOLD SCOPE session spends its 240 s post-selection
window on model work: 37176837432 skipped the handoff helper and paraphrased
the mode, 37198445662 ran it, then wrote the plan, loaded sections and probed
Codex for 114 s before the posture chat. The handoff line is the posture.
* B3: accept wrapped unchanged rows in the cropped Edit preview
PR run 37176835584 timed out with the floor's owned Edit menu on screen:
unchanged diff rows soft-wrap under their gutter with a blank marker
column, and a trimmed empty unchanged row keeps only its number. The
parser accepted wrapped +/- rows only, so it never verified the owned
edit. Both forms are now parsed and still compared with the exact source
lines; the captured viewport is replayed with negative controls.
* A3: record per-phase headroom diagnosis for docsync, plan-mode-no-op, mode-routing and gate slice 11
* E4: /review Step 1 keeps going when the base fetch fails and reports stale base coverage
Under Codex's read-only .git, 'git fetch && merge-base && diff --stat'
stopped at the fetch, so the empty output read as 'Nothing to review'.
The fetch is now separate, prints BASE_REFRESH fresh or stale with the
local base revision, and the review states stale coverage.
* D2: reword the format-approach question prompt
Both plan-ceo-review-format-approach refusals in census 37195203538
(Fable 5.1 model_refusal_no_fallback, category reasoning_extraction:
"duplicating model outputs") were anchored on the session prompt, whose
capture instruction asked for the "verbatim" question and "the exact
prose you would have shown" in a "format-capture test". The approach
case now asks for the same question in the same file as the step's
output, without that wording. The other three plan-format cases keep
their unchanged prompt (no refusals in 72 trials).
* C2: /qa defines its outcome filename fields; QA report templates name the measured timing fields
Low judge samples in failing panels named both: qa workflow (37186854666)
'an undefined {user} in the outcome filename', and qa-only workflow
(37198445662) 'template asks for Duration / durationMs totals while
reporting.md mandates Probe budget and Guarded command time'. The qa
union stays under its 1.11 size ratio (93292 of 93294 bytes).
* 0.1: session ledger written by the claude -p, Agent SDK, PTY and Codex runners
Each armed model session appends one best-effort line to
$GSTACK_EVAL_DIR/session-ledger.jsonl: a stable session key, the timeout it
armed, monotonic elapsed time, its end reason (completed, session_timeout,
observer_timeout, api_error, refusal, aborted, error), one sanitized evidence
line and, for stream-JSON runners, a compact liveness summary (longest
in-request silence with nothing outstanding, last event kind, tool open at
end). Partial messages now stream for every claude -p session and for SDK
sessions without a custom query provider (runSharedInteractive opts in); their
rows feed only the summary, so transcripts keep their size unless
publicStreamDiagnostics is set. Structured refusal and API-error signals win
over phrases, and tool results are never read for them.
* 0.2: failure_cause beside failure_class, from runner facts and the session ledger
One pure failureCauseOf() beside failureClassOf() sets failure_cause
(contract > pre_turn_infra > api_error > refusal > provider_stall >
session_timeout > observer_timeout > assertion > unknown; no machine-set
'detector') with one bounded evidence line, and failureDetailOf() keeps a
Bun matcher's Expected/Received or a judge's failing dimensions in
failure_detail. Both record builders (classifyTrialShard, junitCensus) call
them and attach the trial's ledger sessions. failure_class, error and
TRIAL_ERROR_MAX are computed exactly as before; the new fields are optional,
bounded and validated only when present, and a frozen copy of the v1 reader
accepts every new record. A provider stall needs partial messages and
STALL_WINDOW_MS (120 s, eval-budgets.ts beside HEADROOM_WARN/FAIL) of
in-request silence with nothing outstanding; it stays a failure under
EVAL_POLICY v1.
* E4: fingerprint under a read-only .git via a private object dir owned by review-log
Codex's sandbox mounts .git read-only, so gstack-wtree's git add -A could not
write blobs and /review lost its start capture. gstack-review-log now creates
a private 0700 object directory under TMPDIR, passes it to gstack-wtree (which
writes new objects there with the common-dir object store, resolved before
exporting, as an alternate, so linked worktrees work), exposes it to the
bindReview/checkSharedLibsReuse consumers that read the tree and blobs, and
removes it on EXIT/INT/TERM after they finish. gstack-wtree prints one stderr
line and exits 1 on a failed write or unwritable TMPDIR instead of printing a
partial hash; review-log no longer swallows that line. Standalone gstack-wtree
keeps writing to the repository store so git diff between saved snapshots
(/ship) still works.
* 0.3: red lines name their values, evidence and an after-a-repair command; one published-text sanitizer
Each red line in the census report keeps its stored error and adds the
machine cause with its evidence, the failure_detail values (Expected/Received,
or each failing judge dimension with its mean, sample count and the sample
sentence naming it), the slice artifact and shard directory that hold the
transcript, a fetch command (eval:pass-rates --run <id> --case <id>), and
'after a repair:' with a command that selects only that case (E2E ids via
--case, judges via their selection and an exact test-name filter). A
session that completed and failed a check says so instead of 'assertion
(success)'. Report text that reaches a PR comment, issue or committed file
goes through scripts/lib/published-text.ts: control and bidi characters
stripped, credential-scan patterns redacted, mentions neutralized, HTML and
link brackets escaped inline, blocks in a fence longer than any backtick run,
length capped. Verdicts and collector verdict objects are unchanged.
* E2: one make-pdf resolver for readiness and every $P block on env-var hosts
make-pdf exported $P from its readiness block, which Codex and the other
env-var hosts lose between blocks. binaryAssignment gains 'make-pdf': the
MAKE-PDF SETUP readiness check and the runtime prelude both use it, so every
block that runs $P derives the same make-pdf/dist/pdf binary, with
MAKE_PDF_BIN (dev builds) winning when it is executable. Claude output keeps
its repo-local-first probe; only the false 'subsequent blocks' comment
changed. Root plus P=... fits the 400-byte prelude budget; the fixed
MAKE_PDF_OVERRIDE segment is the one exemption, named in runtime-root.ts and
the conformance test.
* 0.1: eval:pass-rates --headroom, --reds and --run report views; 85% headroom alarm in --gate
A new pure module, scripts/lib/eval-history.ts, reports per case the slowest
armed session against the timeout it armed (censored when it timed out,
insufficient below 3 samples, unknown without a session ledger, with the
case wall shown separately as an upper bound), the census critical path from
the slice jobs' timings, the verdict red ledger per census by failure class
and machine cause with shrunk per-case rates, and the all-green probability
(an approximation: the product of 1 - p_i, with an interval). --run <id>
lists one census's reds with their values and history and downloads only the
slice artifacts those reds name, reporting any over 64 MB instead of skipping
them. eval-flake-rank.ts only routes flags: --help works offline, an unknown
flag or case id exits 2 before any fetch, --case narrows the JSON and the
alarms, every view prints its run/branch scope, and --gate also alarms on a
session above HEADROOM_FAIL (85%) of its budget.
* E5: GSTACK_DESIGN_IMAGE_MODEL override; the live model check always tests the defaults
The image_generation tool model gains an env override validated against a
conservative gpt-image-<version> pattern; a bad value fails before any
request and the error names the variable. Pairing and rejection hints name
the tool model actually requested. design/scripts/live-model-check.ts now
ignores GSTACK_DESIGN_MODEL and GSTACK_DESIGN_IMAGE_MODEL (and says so), so
an override cannot mask a retirement of a default, and its exit 2 (no key)
says nothing was checked.
* C3: /deslop-shared-libs names a concrete risk for each recommendation, never 'none'
Both shared-libs-opportunity-judgment reds (37193478719 t3, 37195203538 t1)
failed adoption_and_risk because the top recommendation said 'Main risk:
none found inside the repo' / 'none in behavior'. The skill asked to 'name
the main risk or uncertainty' without saying what counts.
* B4: --case selects loop-registered cases; an unknown id exits before any process
caseSelection() resolves the one paid file that registers a case and how a
trial selects it: by its literal Bun test name, by the id a registration loop
names its test at runtime (coverage audits, TPA, journeys), or by running the
whole file when it registers no other case (plan-mode-no-op, whose test names
are not the id, so the old name pattern ran nothing). A whole-file case runs
one file shard per trial and is judged by shard status, as CI judges it.
--list prints the selection. Every E2E_TIERS id now resolves except
carve-section-loading, which fifteen files register (CI runs each as its own
shard) and which still fails with the owning files named.
* 0.3: after-a-repair runs the shard file when --case cannot resolve a case
A case that several files register (carve-section-loading) cannot be selected
by --case; its red line now prints the bun test command for the shard file
that went red, so every printed command selects exactly one case.
* C1: pre-register workflow-judge control bundles for qa, qa-only, review and setup-deploy
Positive (production bundle), negative (headings only) and one targeted
execution-blocking defect per judge, one 3-sample panel each, built with the
eval's own reader, prompt builder, options and floors.
* F: cost_known: a trial whose harness captured no billing says so; totals never hide it
cost_usd stays a finite number. Trial and eval records gain an optional
cost_known boolean (absent means known, so old records still parse; the
validator accepts only a boolean). Session-ledger rows record whether the
session's terminal result carried total_cost_usd; PTY and Codex sessions
never do, and SDK captures without a terminal result do not. A trial is
cost-known only when it has eval records and every record and session
captured billing, so JUnit-only PTY cases (125 of 298 trials in run
37198445662 recorded 0) now read cost unknown. The census headline prints the
known sum plus the unknown count, never a bare total over unknown costs, and
no total when every cost is unknown; the history import keeps the flag.
PTY transcripts are not priced: test/helpers/pricing.ts estimates exclude
cache writes, so they cannot give the billed cost.
* B1: replay corpora for the five repeat-offender cases
shared-libs-plan-callers, plan-ceo-section-loading, auto-decide-preserved,
plan-design-review-plan-mode and ship-docsync-late-result each get an
inventory of every wave census in scope (all artifacts still available)
and minimal, redacted detector or actor inputs replayed through the real
detector in the free suite: census passes, every distinct census red, and
derived negative controls labelled as derived. Fixed false reds stay as
expected passes with their history; the two genuine auto-decide product
failures (37176837432, 37179171083) stay refused. A contract test bounds
each entry at 64 KB, scans it for HIGH credentials, requires a pass and a
refusal per case, and reports entries whose series identity is stale.
Building the auto-decide corpus exposed a latent native detector misread:
a bold 'Mode application and rationale.' heading was parsed as a
withdrawn Mode field (37186854666). A field status now ends at its
sentence.
* B2: read approved stale-fill decisions from more valid ledger records
The structured ledger reader is authoritative before the prose fallback,
but it missed two valid approved records in passing censuses
(37186854666: a camelCase writeProfile subject; 37195203538: a
slash-joined 'D1 / STALE-FILL' row id), which then rested on the prose
detector that caused both historical reds. Both forms are accepted; an
unresolved row, an unrelated decision, the copied fixture plan and
non-write camelCase words stay refused.
* C1: record workflow-judge control results (negatives separated; qa and review miss a removed step)
* C2: record every low workflow-judge sample in the wave censuses with its disposition
* 0.3: published lines cannot break a fixed fence
evals.yml wraps collector-outcomes failures in a fixed three-backtick fence
for the PR comment, so those lines (and eval:pass-rates output meant for the
tracking issue) go through sanitizeFixedFenceLine: cleaned like every
published line, with backticks turned into quotes. Inline published text
also neutralizes backticks.
* A1: fit the post-child record wording under ship's 1.428 parity size cap
* E4: inline the Step 1 base refresh to fit review's 1.199 parity size cap
* B2: inventory of gate and periodic phrase-only detector checks
docs/evals/detector-inventory.md lists every pass/fail decision in gate
and periodic E2E cases that rests on phrases in model free text, with
its structured source (if any), whether the contract is user-visible,
and a disposition: already structured-first, convert candidate, phrase
witness required, or no structured source. Convert candidates are
ordered by census red history.
* 0.2: failureCauseOf takes the stall window instead of importing it; the PTY ledger never throws
eval-store.ts imported STALL_WINDOW_MS from eval-budgets.ts, which free tests
replace with mock.module stubs (review-finalization-budget, office-hours,
qa-fix-loop, qa-only), so loading eval-store failed there. failureCauseOf is
now dependency-free: the record builders pass STALL_WINDOW_MS. A PTY session
whose handle cannot show its text (test doubles) records no panel instead of
throwing from runPtySession's cleanup. The 37186854666 replay also runs
through classifyTrialShard.
* A2/auto-decide: the CEO handoff chat repeats the helper's printed line verbatim
Claude Code 2.1.284 collapses the helper's Bash output ('Ran 1 shell
command'), so the 'Auto-decided review mode → …' line reaches the user only
through chat. In lane D's auto-decide-preserved corpus every passing capture's
chat starts with that line, and the reds paraphrased it ('HOLD SCOPE was
auto-selected…', 'Mode decision done…', "I'm defaulting to HOLD SCOPE…").
The template now says verbatim and why.
* C2: record the judge panels rerun on the final ship and review wording
* integrate: touchfiles and module-size ratchet entries for the wave's new modules
GLOBAL_TOUCHFILES gains session-ledger.ts, published-text.ts and
auq-substance-panel.ts; bin/gstack-bun-version.sh joins every entry that
lists setup or the updater; auq-matrix stays a rule case with its C4
evidence beside E2E_KINDS; eval-history, published-text and session-ledger
are registered new modules.
* E1: BROWSE SETUP install hint pins the tested Bun (1.4.0)
* E2: quote "$P" in make-pdf commands so install paths with spaces work
* E1: /gstack-upgrade checks the incoming Bun floor before pulling
gstack_bun_incoming_hold (bin/gstack-bun-version.sh) is now the one
pre-advance check: the auto-updater calls it before its fast-forward and
/gstack-upgrade's git path calls it after fetching. Below the incoming
release's floor the upgrade stops with BUN_TOO_OLD and the checkout,
VERSION and installs stay untouched. Vendored installs were already safe:
setup refuses before writing and the swap restores the previous install.
* 0.3: only main posts or closes tracking issues; reports carry alarms, red ledger, headroom and all-green
A branch-dispatched periodic or marathon run writes its report to the step
summary and a report artifact; the upsert and close steps run only on
refs/heads/main. The periodic body now keeps every pass-rate alarm line (the
old grep kept only the header), adds the red ledger and session headroom read
with an explicit --branch, the all-green estimate and the triage guide link,
and routes report text through scripts/lib/published-text.ts. The marathon
lane gains the matching close-on-green step.
Migrated: evals-workflow-wiring pinned report-summary.md in the upsert step;
the body is now written by 'Write the census report'.
* C3: prosons neutral-posture detector ignores a negated "taste call"
Both plan-review-prosons-neutral-neg reds (37179171083 t3, 37193478719 t3)
recommended A and closed with "a coverage call, not a taste call"; the
old /taste call/i matched the negation. The detector moves to a helper so
the two census captures replay free as negatives, with positive controls.
* B2: credit the CEO mode handoff only when the chat shows its line
Claude Code 2.1.284 collapses the gstack-ceo-mode-handoff Bash result, so
the printed AUTO_DECIDE line is user-visible only when a later chat message
of the session begins with it. The native detector now requires that
witness; the helper's tool result locates it but no longer replaces it, and
a generic mode statement cannot stand in once a handoff ran. Census
37182865432 is relabelled a genuine miss (expected: refused): the chat said
"I'm defaulting to HOLD SCOPE…" and "Mode decided: HOLD SCOPE". Corpus
replay: 5 credited, 3 refused. ceo-mode-handoff tests migrated to the
witness rule (deliberate behavior change).
* test: baseline the wave's approved detector and harness replay tests
B1 corpora, B3, C3, C4 and D2 replay captures through paid-case detectors,
actors or fixtures (test code by design); each entry carries its reason.
* E6: live Codex multi-block periodic case in the host-run Codex job
test/codex-e2e-multiblock-live.test.ts runs ./setup --host codex once from the
job's checkout into a fresh HOME and CODEX_HOME (auth.json copied from the CI
login home), asserts codex debug prompt-input lists the gstack router as a
real file, seeds one learning, and runs /learn's Show recent block through
codex exec: that later block must root itself through its prelude, run the
installed gstack-learnings-search and print the seeded learning; no block may
report a missing install or an unresolved gstack helper path.
The file joins CODEX_CI_FILES so the planner gives it to the host-run Codex
job. paid-test-durations.json carries a 180 s packing estimate (refreshed by
--write-durations after the first census) so the plan keeps every slice
within max-parallel. Oracles live in test/helpers/codex-multiblock.ts with
free replays in test/codex-multiblock.test.ts. Registration (E2E_TOUCHFILES,
E2E_TIERS periodic, E2E_KINDS rule) is the integrator's.
* E5: design live-model check as a periodic case
test/skill-e2e-design-model-smoke.test.ts runs design/scripts/live-model-check.ts
(default models only) under a temp HOME. Without OPENAI_API_KEY the case is a
Bun skip that the census lists as SKIPPED with no coverage credit; with a key,
exit 0 passes and any other exit fails with the script output (exit 2 under a
present key fails as 'nothing was checked'). paid-test-durations.json carries
a 90 s packing estimate. Registration (E2E_TOUCHFILES, E2E_TIERS periodic,
E2E_KINDS rule) is the integrator's.
* A1: docsync parent points children at a saved interface file
The /ship docsync fixture saved its observation interface only in the
parent prompt and told the parent to copy it into every child prompt,
about 30-40 s of parent generation per session (lane S, 37198445662).
docsSessionOptions now writes it to <home>/fixture-interface.md (0600) and
the parent tells each child to Read that file first. Fault-actor prompts,
which prepare already writes to a file, are unchanged. Fixture-only.
* A3: a JUnit case's recorded wall is capped by its shard's wall
junitCensus summed a case's JUnit test times as duration_ms; for
plan-mode-no-op's concurrent PTY tests that is 346 s inside a 176 s shard
(the "432-613 s" history). The case wall headroom shows is now
min(sum, shard elapsed), a true upper bound. Session headroom already
used the ledger's per-session clocks.
* A3: refresh paid-test durations from census 37198445662
Recorded 2026-09-29 estimates under-packed gate slice 11 (17 files
estimated 1078 s ran 1353 s; plan-review-report 76 s -> 198 s,
plan-ceo-plan-mode 37 s -> 112 s, safety-codex-consult 83 s -> 150 s).
Refreshed both tiers with --report --write-durations from the latest
census's slice walls. Old plan re-costed with observed walls: worst slice
~677 s (slice 11); refreshed plan: every slice estimated <= 539 s.
* C2: Browser Quick gets a feasible budget; QA loop names its directory and stop path
Judge rationales on the qa-only panels (2 of the last 3) cited the 30 s
Browser Quick budget: homepage + 5 targets, each a checkpoint write and a
guarded command, left ~5 s per probe. Quick is now 180 s (method text: 3
minutes); the qa-only-no-fix paid check expects the same native deadline.
PROBE_DIR points at its source line, a stopped loop goes to the final
report steps (annotations, materialize, Markdown), and /review and /ship
say that after smoke expiry only plan checks rerun in revalidation.
* C2: name the flows judges found implicit in review resolvers
Structured outside review: MISSING COVERAGE asks no fix question and is
never reported clean (was "preserve the existing user decision flow").
/review's HIGH-impact plan gate: spawned or non-interactive runs report
REQUIREMENTS MISSING and continue, writing no TODOs. Value cards: the byte
clamp applies to card lines; JSON records you write keep full values.
* C1: a free test catches a removed workflow step, which the judges missed
C1 controls showed the qa workflow judge passing /qa without Phase 10 and
the review judge passing /review without Step 3. The new test pins every
step/phase heading of qa, qa-only, review, ship, plan-ceo-review,
plan-eng-review and setup-deploy (derived from the templates' step
headings) and names any that goes missing or out of order; new step
headings must be pinned. The judge rubric is unchanged (EVAL_POLICY).
* size: fit the wave's wording under ship, review and qa parity caps
Measured after all lane merges: ship -79, review -210, qa -68 bytes over.
Trimmed the added wording first (post-expiry rule, MISSING COVERAGE,
plan-gate default, value-card JSON, stop line) and folded /review's
redundant INFORMATIONAL sentence and findings bullet. ship 268,036 (room
8), review 130,118 (room 1). qa needed 11 more bytes: cap 1.11 -> 1.1102
with the measured 93,305 / 84,049 = 1.11013 and its reason beside it.
* docs: troubleshooting sections for the wave's new messages
Bun floor (setup, auto-update, /gstack-upgrade), /review stale base and
fingerprint temp-dir errors, the design image model override, the
unrecorded Chromium PID, and a contributor section for census report
lines (pass-rates flags, failure causes, headroom, cost unknown, overlay
v4, stale corpora, auq substance). A free test checks every
troubleshooting.md#anchor printed by setup or bin scripts exists.
* TODOS: reliability follow-ups deferrals; close the Bun floor, make-pdf $P and #2709 reap
* 0.4: 'A census went red' guide; docs for causes, session ledger, headroom and new messages
docs/evals/census-red.md walks one red line to a diagnosis: prerequisites, the
report fields, eval:pass-rates --run/--reds/--headroom with expected output,
what unknown/insufficient/unrecorded data means, free replay, and the single
'after a repair' paid run allowed by TESTING_INTERNALS. Linked from
CONTRIBUTING, TESTING_INTERNALS and both census reports.
TESTING_INTERNALS documents the failure_cause glossary (one row per
FAILURE_CAUSES value, diagnostic only under EVAL_POLICY v1), failure_detail,
sessions[], cost_known, the session ledger, the pass-rates views and the 85%
headroom alarm, main-only tracking issues, overlay contract v4, auq-matrix's
3-sample substance panel and the E5/E6 periodic cases. troubleshooting.md
gains #bun-too-old and an anchor per new user-facing message from lanes P, D
and F. README names GSTACK_DESIGN_IMAGE_MODEL. test/census-red-docs.test.ts
pins the glossary to FAILURE_CAUSES, parses every guide command and checks the
links.
* v1.91.20.0: CHANGELOG and version for the reliability follow-ups
* regen: SKILL.md and sections for all hosts, digest, context-budget fixture
bun run gen:skill-docs --host all; goldens and llms.txt re-rendered
unchanged; bun test/helpers/capture-context-budget.ts.
* integrate: register codex-multiblock-live and design-model-smoke (periodic, rule)
* 0.4: census report and eval:pass-rates views link the census-red guide
* test: cookie approval fixture reverses the Bun install-hint pin (E1)
The historical approved cookie-workflow prompt carried BUN_VERSION=1.3.10;
the regenerated install hint now pins 1.4.0, so the fixture's reversal
list gains that edit like the earlier waves' entries.
* test: release floor counts the two new periodic cases (E5, E6)
* fix: /freeze, /guard, /unfreeze and /investigate work on Codex and other env-var hosts
Templates quoted "$HOME/.claude/skills/gstack/...", which the env-var hosts'
rewrites turned into "$HOME/.agents/skills/gstack/..." on Codex (a path a
global install never has; same class on the other hosts), and setup's
runtime roots had no freeze/bin. Quoted runtime paths now use
{{RUNTIME_ROOT}}: $GSTACK_ROOT on env-var hosts (the fence gets the
runtime prelude), the unchanged $HOME path on Claude. _link_runtime_dists
links freeze/bin and the careful/bin helper it sources into every
env-var host's root. Also fixes the same guess in /benchmark-models,
/retro and /diagram. test/env-host-runtime-assets.test.ts fails on the
previous code: no fence guesses a $HOME/<dir>/skills/gstack path, every
$GSTACK_ROOT path a fence uses is in the host's runtime root, and the
installed Codex /freeze and /unfreeze fences run from a fresh shell.
Template-reading tests now read the Claude render or expect the
placeholder (deliberate change).
* CHANGELOG/TODOS: lane W's workflow, cases and guide; the env-var host freeze fix
* E6: a block that reads a SKILL.md is not a missing install
The first live codex-multiblock-live run passed its real check (the /learn
block found the install and printed the seeded learning) but the oracle
flagged a sed of the learn SKILL.md, whose prelude source quotes the
'gstack: no install found (tried $_r)' text. Only an emitted line that
starts with the message and names a real root counts now; free replay
negative added.
* test: qa-only fixture keeps its caller-set 30 s deadline
The qa-only-no-fix fixture tells the model to start the guard with 30 s
(a caller limit), so its deadline assertion's default stays 30 000 ms; the
product's Browser Quick default (180 s) does not apply to it. Reverts the
helper and its test to their base.
* C2: keep PROBE_DIR 'owned' (qa-functional-prompt pins it); qa measured 93,311
* test: paid file count and periodic floor include the two new periodic cases
* test: ship docs freshness pin checks meaning after A1's wording trim
The snapshot still hashes selected release paths (--select narrows them),
generated outputs and docs; the pin now checks that instead of the
pre-trim sentence.
* E3: reap the recorded Chromium only after the daemon has exited
With the PID now recorded (E3), browse stop reaped Chromium while the
daemon was still shutting down; the daemon read that as a crash, exited 1
and left its state file (chromium-profile-isolation's headless stop test
failed 3/3 locally, passes on base). The CLI waits up to 5 s for the daemon
to exit before reaping a survivor.
* regen: qa and qa-only exploratory sections (PROBE_DIR wording)
* B1: batching counter reads a target declared mid-line; replay corpus
GSTA-23's census 37228573062 failed plan-eng-multi-finding-batching with
reviewCount=0 after 975 s although D1-D10 were asked one per finding,
each with its saved decision-ledger record. The report wrote 'Report
destination requested by the user. Review target (fixed): `PLAN.md` ...'
and the counter read target fields only at a line start, so no brief
inherited the plan identity. A target field may now open any sentence of
its line; a mid-sentence mention or another file still gives no identity.
Floor (3), budget and kind unchanged. Corpus: 9 earlier passing censuses,
the red (now credited at D5), and a derived batching refusal; 975 s noted
in the headroom diagnosis.
* B1: shared-libs corpus labels its git path and covers the supported shape
Every census entry was recorded in CI's eval image (git 2.43), where
gstack-safe-git falls back because --no-lazy-fetch is rejected; each entry
now says so (git_path: fallback). Two synthetic supported-path entries
(HEAD and a history citation in the brief, as git >= 2.44 produces) keep
the 37193478719 pass and the 37195203538 refusal, and a test asserts both
shapes are covered and the actor reads no git-path-specific text.
* v1.91.21.0: move off v1.91.20.0 (claimed by #3031)
* test: settle nav-guard commands before asserting; make two new tests Windows-safe
navigation-guard: Bun's expect(promise).rejects slows each Playwright round-trip
to ~1s while it waits (50 page.evaluate: 43ms plain vs 46.6s inside the matcher),
so the click's 5s timeout expired before the guard saw the link-local request
(CI: 'click: Timeout 5000ms exceeded ... waiting for scheduled navigations').
Settle the command first, then assert on the refusal; a source check keeps it so.
env-host-runtime-assets: PROBE_FILES now matches either path separator.
detector-corpus ship-docsync-late-result: JSON-escape the <ROOT> substitution so a
Windows temp path's backslashes do not break JSON.parse.
* fix: CI run/job ids are not phone numbers; context-recovery fixture gets the full runtime bin
pii.phone.e164 redacted the run id in the published eval report's fetch command
(--run 37235771700), breaking the command it prints. Digit-only spans in a run/job
flag, gh run argument, actions/runs/ or /job/ URL segment, or run_id/job_id key are
now exempt; the same digits elsewhere still report.
context-recovery-artifacts: the fixture's ./bin/ lacked gstack-skill-start and
gstack-context-recovery (the preamble's runtime scripts since v1.71), so the agent
spent its 10 turns hunting for them and hit error_max_turns with no final message
(run 37235771700, 11 turns). Link the whole runtime bin; diagnostic rerun passed
in 6 turns.
* test: replay run 37237194905's late-result capture through the docs authority detector
The only failure was call 20's write to a mistyped shard root (3GanyY for 3GanyU),
which the model disclosed; the detector is correct and the A1 post-snapshot
--out under the fixture home stays permitted. Named red recorded in TODOS.
* test: keep docsync-authority in the Windows profile
The replay test's literal helper path tripped the Windows curator's shebang-spawn
pattern, dropping the whole file from the Windows profile (test-free-shards
invariant). The snapshot --out case is already covered by docsync-command-grammar,
so it is removed here, and the Linux-path replay skips on win32.
* fix: gstack-docs-candidate --help works; docs gate red lines name the denying tool failure
PR run 37239016858 ship-docsync-completion: the parent ran
'gstack-docs-candidate --help' (the helper's usage error points at --help, which
failed). That undeclared command made docsToolFailures non-empty, which denied
the atomic-replacement proof of the child's legitimate native Edit, so the red
line blamed the Edit's temp file. A replay of the capture proves the atomic
replace once --help is declared.
- the helper prints its usage for --help/-h and exits 0
- the fixture interface and command grammar admit the helper's bare --help
- the atomic-proof denial names its check and the tool failures behind it
(deny labels renumbered to their lines)
- replay with negative controls (undeclared command, temp left behind, renamed
elsewhere); diagnostic rerun of ship-docsync-completion passed
* test: retro-base-branch fixture uses main and a shallow report path
PR run 37241493377: the agent wrote retro-output.md one directory above the
requested <tmp>/retro-base/retro-output.md (it dropped the retro-base segment),
so the report check found nothing. The fixture also ran plain git init, which
makes master on CI's git while the prompt says the repo is on main; the agent
logged that mismatch as a learning. Init on main and use the temp dir itself as
the repo. Diagnostic rerun passed (18 turns, 203s).
* test: qa-quick sets a caller probe limit and drains past its session budget
C2 raised Browser Quick's native budget to 180 s. qa-quick is the only gate case
on the native budget, and its wall went 194 s (main periodic, pre-C2) -> 254 s and
274 s (this PR's runs) -> 300 s session timeout (run 37242691913): the agent now
probes every navigation target with a checkpoint, then needs ~80 s to report.
Like qa-only-no-fix, the eval now sets a caller probe limit (30 s), scopes the
fixture, and gives bun's outer timeout the session drain margin instead of racing
the 300 s session. The product's 180 s budget and the session budget are unchanged.
Diagnostic rerun passed in 195 s.
* review/ship: use main's post-expiry smoke sentence; v1.91.24.0
GSTA-23's measurement on the review workflow judge: main's 'Post-expiry smoke
rechecks are not-run.' passed 10/10; the C2 wording passed 6/10 and the
'..., even in 4c.' variant failed in PR run 37248650879. qa-probe-gates pins
main's sentence before step 4c.
* test: record the plan-design corpus test in the test-of-test baseline
#3031's ratchet now counts tests that slice a paid test file's source; the B1
plan-design-review-plan-mode replay corpus reads test/skill-e2e-design.test.ts
like its sibling corpora already listed.
* fix: --case no longer deadlocks importing the running CLI
runCaseDiagnosis loaded runPaidShards with a dynamic import of
scripts/test-paid-shards.ts; when that file is the entry it is still in its
top-level await, so the import never resolved and `--case` hung after its
first line (on main since #3031, found by the per-case measurement runs). The
CLI now passes its runner in; library callers keep the lazy import.
* test: name the in-flight file and the native error when bun dies before its summary (#3037)
* Bun 1.4.2 everywhere CI and setup pin the tested version
Bun 1.4.0 on Windows aborts about 3% of free-suite shards with native
bad-handle errors (main too; #3037); the same head on 1.4.2 had none in 180
shard runs. Every current 1.4.0 pin moves together, as bun-version-drift
requires: setup-bun in every workflow (CSO orchestration included), the eval
image (sha256 from bun-v1.4.2's SHASUMS256.txt, re-hashed locally), GitLab,
the Ubicloud free-suite setup, gstack-bun-version.sh's tested version,
engines.bun and @types/bun, the install hint in generated skills, the Dia and
cookie native-qualification pins, and docs and tests that state the current
pin. Recorded 1.4.0 observations, fixtures, CHANGELOG history and the CSO
images' oven/bun:1.3.10 base stay.
* fix: --case runs a single-case file whole
A session option naming the case (testName: 'office-hours-section-loading')
made caseSelection pick test-name mode, whose filter matched no Bun test, so
every trial would run hollow. A file that registers only this case now runs
whole before any name match is tried.
* fix(office-hours): reviewer returns a hash-bound receipt instead of echoing its verdict
The spec reviewer now Writes its verdict once, seals it with
gstack-office-hours-review seal (validates the saved JSON and prints
OFFICE_HOURS_VERDICT round=N sha256=<hash> path=<verdict>), and returns
only that line. check --receipt recomputes the hash and fails closed on a
missing, malformed, stale-round, wrong-path or mismatched receipt. The
saved file stays the sole findings inventory; validators bind the receipt
to the saved bytes and the judge reads those bytes.
* fix(office-hours): only blocking findings send a design back for another review round
Review findings carry severity blocking|minor (verdict schema v2). PASS when
no blocking findings remain; minor findings are recorded in Reviewer
Concerns and the Spec Review table, never a reason for another round.
CONVERGENCE needs a persisting blocking prior. A persisting or unverified
blocking prior cannot be relabeled minor. The 3-round cap is unchanged.
* fix(office-hours): rounds 2 and 3 are delta re-reviews of the captured design diff
prepare saves each reviewed design as round-N.design.md and gives later
reviewers the exact line diff since the previous round. A new later-round
blocking finding must cite (changed_text) an excerpt of a changed line;
seal, check and the validators verify it against the snapshots and fail
closed when a citation or snapshot is missing. Other new observations are
minor. Severity, anti-downgrade and CONVERGENCE rules are unchanged.
* test(office-hours): preservation judge accepts COMPLETED with recorded minor findings
* fix(office-hours): parent fixes only blocking findings; callers may set --max-rounds
Between rounds the parent edits for blocking findings only; minor findings
stay recorded. gstack-office-hours-review takes --max-rounds 1|2|3 (default
3) on every command; reaching it stops at MAX_ITERATIONS. The marathon
fixture sets 2 rounds so round 1 (full) and round 2 (delta) run live.
* docs(changelog): office-hours review loop changes in v1.91.28.0
* fix(cso): helper JSON keeps run URLs only in the validated workflow slot
The shared redactor now treats GitHub Actions run ids as CI metadata
(c245d1bc2), so main's #3041 invariant that the same run digits stay
redacted outside qualification.workflow held only through the phone
false positive. Make it explicit in the CSO sanitizer.
* test(plan-ceo): split-overflow progress detector accepts the skill's candidate wording
Collection never ended when the model wrote 'E1 Slack:' / 'E1-SLACK:' or
labels like 'Include in this quarter' / 'Defer with flip trigger', so the
session ran the whole review (930-1464 s) and failed on its extra findings
or the case deadline (6 of 21 runs). The candidate check now accepts a
platform-bound E-id and the action is the bucket's leading verb, as the
skill names them (Include/Defer/Cut/Hold). The semantic judge, call floor
and include/defer/cut validation are unchanged. Replays the captured CI
37374526437, t2, t9 and t15 menus with negative controls.
* test(plan-ceo): split-overflow candidate identity is the one E1-E5 token
The skill fixes no header or question wording, and shape-matching kept
missing new forms (header 'D1.1 Slack' with 'SCOPE-E1:', 'D1.1 E1 Slack',
'E1 Slack DM bot ...:', a misspelled 'Mattermst'). Identity is now exactly
one E1-E5 token across the header and the question's first line, with no
other candidate's platform named. Options rules, judge, call floor and
include/defer/cut validation are unchanged. Replays all 31 measured runs;
the quoted-row controls now quote a second ledger id.
* v1.91.30.0: move past #3044's claimed 1.91.29.0
* test: register the split-overflow wording replay as a free fixture
* fix(skills): resolve judge-cited ambiguities in /qa and /plan-eng-review
- tasks JSONL keeps the raw branch as GIT_BRANCH instead of overwriting
gstack-slug's sanitized BRANCH used for artifact names
- the Codex outside-voice block runs for ready or unverified, matching
its routing rule
- plan-eng startup names the preamble's own setup prompts; the
cross-project question waits for Step 0
- /qa 8c: a fix that applies but leaves the finding reproducing gets one
revised fix, else it is reverted; browser verification names its before
screenshot; the exploratory loop drops its duplicate 3-step summary
* fix(plan-eng): keep the startup steps in their pinned order
* fix(plan-eng): explicit spawned sessions follow the spawned rule at the scope gate; one announcement rule
A missing tool still sends the prose menu whatever the session type, but a
session the caller or host explicitly marks spawned selects the menu's
Recommendation without prose or BLOCKED. The pre-load announcement
sentence now states Step 0's single check.
* eval policy v3: judge panels gate on the per-dimension median of 3 samples
A judge dimension now passes when at least 2 of its 3 samples meet the
unchanged threshold (judgePanelMedian); any other sample count or a
non-numeric sample fails closed. Thresholds, sample count, judge model and
prompts are unchanged; panel logs show every sample and the mean, and name
the median as the gate. Applies to the skill-llm-eval panels, the
workflow-judge cache, the auq-matrix substance panel and the calibration
replicas. EVAL_POLICY 2 -> 3 and HARNESS_VERSION 1 -> 2 (verdicts change),
so every case starts a new pass-rate series. Replays the captured panels
from jobs 112032744416 and 112049856586 as replayed evidence.
366 lines
40 KiB
JSON
366 lines
40 KiB
JSON
{
|
||
"recorded_at": "2026-10-04T17:58:09.689Z",
|
||
"results": [
|
||
{
|
||
"judge": "qa/SKILL.md workflow",
|
||
"bundle": "positive",
|
||
"verdict": "pass",
|
||
"mean": {
|
||
"clarity": 4,
|
||
"completeness": 4.333333333333333,
|
||
"actionability": 4
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The bundle has a strong skeleton: a section index with explicit load conditions, repeated STOP/Read directives with a uniform missing-asset blocked path, a well-defined browser-vs-functional split (scope.md surface selection, isolation rules, LOCAL host definition), an explicit caller-authority matrix for browser setup (report-only vs standalone, unknown → report-only), a concrete functional contract map with enumerated outcomes, and a fully specified fix loop (8a→8f) with regression-before-repair, value card, commit/classify/revert rules, self-regulation triggers and a hard cap. Tooling is executable (bash scripts, DEADLINE_TOOL/EVIDENCE_TOOL invocations, health-score weights and rounding, both report templates, screenshot naming). Weaknesses that keep it from 5s: prose is extremely compressed and sometimes terse to the point of ambiguity (e.g., checkpoint 'exploration-NNN.json' rules, 'Guard if bounded; await completion', 'preserve every safe program-JSON key/value and identity hash'); the interplay between exploratory.md's loop and qa-patterns' Phases 1-6 'as techniques' requires careful reading; there is mild tension between 'extend an existing table-driven test' and 'use new native test files / {name}.regression-N.test.{ext}'; functional report naming ('safe command/service label') is vaguer than the browser convention; and some execution depends on assets not in the bundle (browse/SKILL.md Cookbook, /devex-review, gstack tool semantics). An agent can execute and produce the deliverables, but must work hard to parse the dense rules."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The bundle is well-structured: SKILL.md gives an ordered phase sequence (Setup → learnings/test-plan context → baseline probes → triage → fix loop with regression-before-repair → final QA → report → TODOS → learnings) with explicit, conditional Read directives and defined behavior for missing/unreadable sections (block affected probes, continue safe checks). Authority (report-only vs standalone, caller-owned decisions), isolation (LOCAL vs non-local, owned fixtures, one AskUserQuestion for mutations), native contract map, deadline/evidence tooling, checkpoint publication, severity rubric, output naming, templates and taxonomy are all specified with concrete bash/aside/$B commands, making it largely executable. Weaknesses: exploratory.md is extremely dense and terse (e.g., 'Preserve every safe program-JSON key/value and identity hash unchanged', 'QA_DEADLINE receipts are not observations', PROBE_DIR vs REPORT_DIR selection) and would require careful parsing; the DX path is a single sentence in scope.md with no location for /devex-review; the 8a.5 value-card 'JSON keeps full values' references an undefined JSON artifact; and some cross-skill dependencies (browse cookbook, deadline/evidence tool flags beyond those listed) are external to the bundle. These create minor ambiguities and small gaps rather than blocking defects."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 5,
|
||
"actionability": 4,
|
||
"reasoning": "The entrypoint lays out an unambiguous phase sequence (Setup → Learnings → Test-plan context → baseline Phases 1-6 → Triage → Fix loop 8a-8f → Final QA → Report → TODOS → Learnings) with explicit Read directives, stated load conditions, and a uniform blocked-path rule (missing/unreadable section → report setup blocker, block affected probes, continue safe independent checks). Authority is clearly partitioned by caller (standalone /qa vs report-only), isolation and ownership rules are concrete (LOCAL host definition, mutating-action consent, owned fixtures, unknown ownership blocks), and native functional contracts are enumerated in a contract map with defined outcome states. Browser-only sections are consistently fenced off from functional runs, and templates, naming conventions (regression-N.test.ext, run-YYYYMMDDTHHMMSSZ, screenshot names), tool invocations (deadline/evidence tools, learnings search, report copy), commit formats, value card, classification states, stop rules and hard cap are all specified. Weaknesses that keep it from perfect scores: the prose is extremely dense and relies heavily on cross-file references ('shared loop's gates', 'Phase 5 is the before', 'surface report template'), the checkpoint/exploration-NNN.json publication protocol is strict and easy to get wrong, a few outputs are only implied (exact functional report filename, mixed-run report layout, 'safe hostname' transformation), and there is mild tension between 'Diff-aware is automatic with no URL' in qa-patterns and 'absence of a URL never forces a browser' in SKILL.md (reconciled only by the surface-selection order). An agent can execute and produce the deliverables, but with notable cognitive load."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "qa/SKILL.md workflow",
|
||
"bundle": "negative",
|
||
"verdict": "fail",
|
||
"mean": {
|
||
"clarity": 3,
|
||
"completeness": 3.3333333333333335,
|
||
"actionability": 2.3333333333333335
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 3,
|
||
"completeness": 3,
|
||
"actionability": 2,
|
||
"reasoning": "Only section headings were supplied, not the instruction bodies, so the evaluation is limited to structure. The outline is well-organized: an entrypoint with a section index, numbered phases 1-11 covering baseline, triage, fix loop (diagnose/reproduce, regression-before-repair, fix, re-test, commit, classify, self-regulation), final QA, reporting, TODOS update and learnings capture; conditional sections for browser setup/fallback, exploratory probing, native functional contracts, test bootstrap; plus report templates and an issue taxonomy. However, several things the rubric asks about are not visible at all in the provided material: authority rules, isolation/safety boundaries, blocked paths, how 'Phases 1-6: QA Baseline' maps onto native (non-browser) surfaces when qa-patterns.md is explicitly 'Browser QA methodology', how scope.md (a lone ### fragment, 23 lines) drives the browser/native/mixed decision, and what triggers DX/test-bootstrap loading. SKILL.md starts at line 427, implying missing preceding content. Inserted steps like 8a.5 and 8e.5 hint at patching that may create ordering ambiguity. Without the actual step text, bounded-probe limits, decision criteria, and deliverable formats cannot be confirmed as executable, so actionability is rated conservatively."
|
||
},
|
||
{
|
||
"clarity": 3,
|
||
"completeness": 4,
|
||
"actionability": 3,
|
||
"reasoning": "What is visible is only a heading-level outline of each file, not the step text, so executable decisions cannot be confirmed. Structurally the bundle is strong: a scope/surface selector loaded before setup, a gated browser-setup aside with an explicit headless fallback, a native-tool functional section with a contract map and evidence retention, a shared exploratory probe loop with parent handoff, a full Test→Triage→Fix Loop (diagnose/reproduce, regression-before-repair, re-test, commit, classify, self-regulation STOP) →Final QA→Report→TODOS update→Capture Learnings chain, two report templates, a severity/category taxonomy, and a health-score rubric. Gaps at the outline level: no visible section naming authority/isolation rules or an explicit blocked-path protocol (presumably buried in Setup/Additional Rules/Browser access decision), the section-index load conditions and mode-selection criteria are not exposed, and Prior Learnings/Capture Learnings have no dedicated file. Clarity and actionability are therefore rated as plausible-but-unverified: the phase ordering is unambiguous, but whether each phase gives concrete commands, exit criteria, and decision tests cannot be judged from headings alone."
|
||
},
|
||
{
|
||
"clarity": 3,
|
||
"completeness": 3,
|
||
"actionability": 2,
|
||
"reasoning": "The bundle exposes a coherent skeleton: an entrypoint with Setup, Phases 1-11, a detailed Fix Loop (diagnose/reproduce → regression test before repair → fix → re-test → commit → classify → self-regulation STOP gate), conditional sections for browser setup/fallback, native functional contracts, exploratory probing, and report templates plus a taxonomy. Phase ordering and the test-before-fix discipline are clear at the heading level. However, only headings are provided, so the actual load conditions for each section, authority/isolation rules, the contract map contents, blocked-path handling, and the browser-access decision criteria cannot be verified. Several ambiguities surface even from the outline: SKILL.md 'Phases 1-6' overlaps with qa-patterns' own 'Phase 1-6' and exploratory's '1-4' numbering; scope.md is a lone H3 fragment with no parent context; decimal sub-steps (8a.5, 8e.5) hint at bolted-on logic; and no explicit section names the stated conditions for loading browser vs functional vs mixed surfaces. Without the body text, an agent could follow the phase order but could not reliably make the surface-selection, fallback, or blocked-path decisions or produce evidence-backed deliverables, so actionability is limited."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "qa/SKILL.md workflow",
|
||
"bundle": "defect",
|
||
"verdict": "pass",
|
||
"mean": {
|
||
"clarity": 3.3333333333333335,
|
||
"completeness": 4,
|
||
"actionability": 4
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 3,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The entry point is well-structured: a conditional section index, explicit STOP/Read gates with blocked-path fallbacks, parameter table, tier/mode rules, clean-tree gate, report-dir ownership rules, and a detailed fix loop (8a–8f) with a regression-before-repair value card, classification, self-regulation caps and commit discipline. Authority (report-only vs standalone /qa), isolation (scope.md ownership/production-forwarding rules), native contract map, browser/functional separation, health rubric and both report templates are all present, so an agent can produce the stated deliverables. Clarity suffers from the extremely dense, jargon-heavy exploratory.md (SECONDS/DEADLINE_FILE/PROBE_DIR/checkpoint publication rules compressed into nested caller-specific conditionals), the 'Phases 1-6' label that only maps to browser techniques while functional runs use a contract map, a missing Phase 10 (9 jumps to 11), a `{name}.regression-N.test.{ext}` naming rule that can conflict with native conventions (e.g., pytest), and several references to external skills/cookbooks (browse/SKILL.md, /devex-review, /setup-browser-cookies) not supplied. Decisions are executable but require careful parsing; some edge cases (mixed-run reporting, bounded browser wrapping of aside repl) are only lightly specified."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The workflow has a well-ordered spine (Setup → Learnings → Phases 1-6 baseline → Triage → Fix Loop 8a-8f → Final QA → TODOS → Learnings) with explicit STOP/Read gates, surface selection before setup, strict browser-vs-functional conditional loading, and clear authority (report-only vs standalone, caller-owned fixes) and isolation rules. Blocked-path handling is consistent (missing assets/tools block affected probes, continue safe checks, never substitute curl/tests). Native functional contracts are concrete via the contract map, EVIDENCE_TOOL/DEADLINE_TOOL commands, and the functional report template; browser work has scripts, $B fallback mapping, a scoring rubric, and templates. Deliverables (REPORT_DIR layout, report filenames, baseline.json, screenshots, exploration-NNN.json checkpoints, annotations.json, regression test naming/attribution, commit formats) are specified. Weaknesses: the prose is extremely dense and telegraphic (e.g., checkpoint field rules, SECONDS/mode resolution, 'QA_DEADLINE receipts are not observations'), raising misparse risk; Phase numbering skips Phase 10; some referenced assets (browse Cookbook, command-list, devex-review) are external; and tension between 'extend existing table' and the mandatory new regression-N file naming is only partly resolved. Overall executable but demands careful reading."
|
||
},
|
||
{
|
||
"clarity": 3,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The bundle is thorough: conditional Read directives state precise load conditions ('never for functional-only'), blocked-path handling is consistent (missing assets block affected probes, continue safe checks), scope.md defines surface selection and isolation crisply, system-functional.md gives a concrete contract map and outcome vocabulary, the Health Score Rubric and both report templates are fully specified, and Phase 8 (regression-before-repair, value card, commit, classify, self-regulation caps) is executable. Authority tiers (report-only vs standalone /qa, LOCAL vs non-local mutation consent) are explicit. Weaknesses hurt clarity more than completeness: the exploratory probe loop is dense and partially self-overlapping (a '(1)(2)(3)' summary followed by numbered steps 1-5 with browser/functional branches for checkpoint publication), heavy jargon (DEADLINE_TOOL/EVIDENCE_TOOL/checkpoints/superseded) requires careful parsing, qa-patterns.md and system-functional.md are loaded only indirectly via exploratory.md rather than a SKILL.md STOP, phase numbering jumps from 9 to 11 (no Phase 10), and 'Phases 1-6' is a browser-technique numbering reused as the baseline heading for functional runs. Minor gaps: mixed-run final-report merging and tool fallback behavior if gstack binaries are absent are thinly specified. Overall an agent can execute and produce the expected deliverables, but would need to read slowly to resolve the cross-references."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "qa-only/SKILL.md workflow",
|
||
"bundle": "positive",
|
||
"verdict": "pass",
|
||
"mean": {
|
||
"clarity": 3.6666666666666665,
|
||
"completeness": 4.333333333333333,
|
||
"actionability": 4
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 5,
|
||
"actionability": 4,
|
||
"reasoning": "The bundle is internally consistent across entrypoint and sections: timers (Quick 30s/Full 900s) match qa-patterns' 5-15 minute Full, REPORT_DIR/PROBE_DIR layouts agree between SKILL.md and exploratory.md, and the deferred-charter ordering is explicit. Decision points are thorough—conflicting --quick/--regression, non-empty output dirs, missing assets/baselines, unauthorized learning stores, and report-only authority are all resolved without gaps. Clarity suffers from extreme density and layered cross-skill path references (e.g., 'relative to the installed qa/gstack-qa SKILL.md directory... beside this host's installed caller skill'), and the 'accidental early Read' recovery rule is convoluted. Actionability is strong: exact DEADLINE_TOOL/EVIDENCE_TOOL commands, four-field checkpoint schema, and templates are specified, though the cognitive load of reconciling several sections mid-run is high."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The workflow is internally consistent despite dense cross-referencing: SKILL.md explicitly defers charters/clocks to Run the Selected Checks and tells exploratory §1 to reuse the REPORT_DIR resolved in Prepare Report Artifacts, so the apparent ordering tension resolves cleanly. Decision points are defined (conflicting --quick/--regression → ask; non-empty output dir → fresh run-UTC subdirectory; missing baseline blocks only that surface's regression). Tooling is concrete—exact DEADLINE_TOOL/EVIDENCE_TOOL syntax, four-field exploration-NNN.json schema, Aside/$B scripts, scoring rubric and templates. Weaknesses: heavy jargon and layered 'caller/owned/PROBE_DIR' definitions slow comprehension, and 'one probe plus checks' vs multi-capture scripts leaves slight room for batching ambiguity. Report-only boundaries are stated repeatedly and unambiguously."
|
||
},
|
||
{
|
||
"clarity": 3,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The workflow is internally consistent and decisions are mostly explicit (conflicting --quick/--regression → ask; missing baseline blocks only that surface; mixed order defaults functional→browser; charters Written before any clock). Deliverables are concrete: REPORT_FILE naming, run-dir collision rules, exploration-NNN.json's four fields, annotations/materialize before reporting.md, value cards, project-copy path. Clarity suffers from dense forward cross-references, e.g. 'Select Surfaces' tells the agent to defer charters to a later section whose §1 reuses REPORT_DIR resolved in a section after it, forcing reconstruction of execution order. Exploratory §4's materialize step is functional-centric ('browser-only: evidence []'), leaving browser-run handling implicit until SKILL.md's Output Structure. Browser Quick (30s) with mandatory per-probe checkpoint Writes is defined but practically tight."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "qa-only/SKILL.md workflow",
|
||
"bundle": "negative",
|
||
"verdict": "fail",
|
||
"mean": {
|
||
"clarity": 2,
|
||
"completeness": 1.3333333333333333,
|
||
"actionability": 1
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 1,
|
||
"actionability": 1,
|
||
"reasoning": "The bundle supplies only headings; every file's body (SKILL.md lines 382-596, exploratory.md, reporting.md, scope.md, browser-setup.md, qa-patterns.md, template) is absent, so no actual instructions, decision criteria, or outputs can be followed. Example: 'Select Surfaces and Isolation' and scope.md's 'Select the surface before setup' give no rule for choosing browser/native/mixed targets or isolation. Example: 'Additional Rules (qa-only specific)' and reporting.md name no concrete prohibitions on mutating product/tests/Git nor evidence or timing field definitions, and the report template lists section titles with no field contents. The outline structure is coherent, which earns minimal clarity credit, but an agent cannot execute the workflow or produce a checked report from this content."
|
||
},
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 1,
|
||
"actionability": 1,
|
||
"reasoning": "The supplied bundle contains only headings; no instruction bodies, decision rules, or output contracts are present. For example, \"Select Surfaces and Isolation\" and scope.md's \"Select the surface before setup\" give no criteria for choosing browser vs native functional vs mixed targets, and the Health Score Rubric lists weights (Console 15%, Links 10%) without counting or scoring rules. The outline structure hints at a sensible flow (preparation gate, probe loop, evidence-based reporting, no-mutation rules), but an agent cannot determine what to run, what evidence to retain, or how to populate the report template. Nothing executable can be derived from headings alone."
|
||
},
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 2,
|
||
"actionability": 1,
|
||
"reasoning": "The bundle supplies only headings; no instruction bodies survive, so every step, decision rule, and output contract is undefined. The outline is coherent (surface selection, isolation, conditional browser setup, evidence-based reporting, no-mutation rules) and the section index maps to real files, which earns minimal structural credit. But decisive gaps remain: 'Select Surfaces and Isolation' and scope.md give no criteria for choosing browser vs. native functional vs. mixed targets, and the report-only mutation prohibition exists only as a heading ('Additional Rules') with no enumerated forbidden actions or enforcement check. An agent cannot execute the probe loop, fill the functional report template fields, or verify coverage limits from this material; it would have to invent the workflow."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "qa-only/SKILL.md workflow",
|
||
"bundle": "defect",
|
||
"verdict": "fail",
|
||
"mean": {
|
||
"clarity": 2,
|
||
"completeness": 4,
|
||
"actionability": 2.6666666666666665
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 4,
|
||
"actionability": 3,
|
||
"reasoning": "The bundle is detailed and otherwise coherent, but the entrypoint opens with a decisive contradiction: SKILL.md's first instruction says 'After reporting each bug, fix it in the source and commit the fix,' immediately followed by 'NEVER fix anything or change product tests,' Additional Rule 1 and exploratory §3 forbidding any product/test/Git mutation. An agent cannot resolve this without guessing, and the wrong guess violates the workflow's core report-only guarantee. Elsewhere decisions are well-specified: mode conflicts trigger a user question, REPORT_DIR/REPORT_FILE resolution, mixed-run clock directories, checkpoint field contents, timing fields and the value-card bar are all defined, and the preparation gate, probe loop, finalization order and templates give a complete deliverable path. Minor friction: scope/method Reads live in the qa directory while the skill is qa-only, requiring path resolution."
|
||
},
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 4,
|
||
"actionability": 2,
|
||
"reasoning": "The entrypoint's second line orders \"After reporting each bug, fix it in the source and commit the fix,\" directly contradicting the next bolded sentence (\"NEVER fix anything\"), Additional Rule 1, exploratory §3 and reporting's report-only repair fields. An agent cannot resolve which authority governs, and the wrong resolution violates the workflow's defining constraint. Otherwise the bundle is thorough: surface selection, mode handling, REPORT_DIR/REPORT_FILE derivation, charter-before-probe ordering, deadline/evidence tools, checkpoint schema, templates, value cards and finalization checks are well specified, with explicit blocker handling. Minor friction remains (the Browser Setup Read points at the qa directory while sibling Reads use qa-only), but the fix-and-commit contradiction is the decisive defect for clarity and actionability."
|
||
},
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 4,
|
||
"actionability": 3,
|
||
"reasoning": "The entrypoint's first instruction (\"After reporting each bug, fix it in the source and commit the fix\") directly contradicts the bolded report-only mandate, exploratory §3 and Additional Rule 1, leaving an agent to guess which governs the core safety boundary. Otherwise the bundle is thorough: surface selection, mode/baseline resolution, REPORT_DIR/REPORT_FILE rules, deadline/evidence tooling, checkpoint schema, health rubric and finalization checks are all specified. A second ambiguity: SKILL.md tells functional-only runs to skip Browser Setup, yet 'Run the Selected Checks' says to read TODOS.md 'before browser probes' without clarifying functional preflight ordering. Deliverables are achievable, but the opening contradiction risks a prohibited commit."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "review/SKILL.md workflow",
|
||
"bundle": "positive",
|
||
"verdict": "pass",
|
||
"mean": {
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The skeleton is dense but each path resolves: Step 4 reads QA scope/methods read-only, Step 4.7 runs bounded smoke even for small diffs without plan/server, test_stub forces ASK (Step 5a), incomplete/blocked probes keep COMPLETED=false, and Step 5.8 reruns affected probes with a new REVIEW_START. Two rough edges keep it from 5: exploratory.md says `materialize` runs once per PROBE_DIR while Step 4.7/5.8 reuse REPORT_DIR and require revalidation captures after repairs, leaving where rerun evidence is materialized undefined; Step 3.4 gives no handling when BRANCH_VERSION or NEXT_SLOT is empty before the version comparison. Outputs (report, log record, QA section) are otherwise well specified."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The skeleton plus on-demand sections give a determinable next action on nearly every path: QA Reads resolved from the installed QA directory, smoke bounded via DEADLINE_FILE while plan checks/revalidation use --timeout-ms, test_stub findings forced to ASK, incomplete/blocked probes mapped to COMPLETED=false, and repeat passes rerunning only affected probes. Minor tensions remain: SKILL.md Step 4.7 step 3 says 'post-expiry smoke rechecks are not-run' while step 4c and Step 5.8 require rerunning affected probes after fixes, leaving which rule governs an expired-smoke revalidation to inference; Step 3 defines a pass as 'before any fixes' while Step 5.8 says a pass covers Steps 3–5. Outputs, records and gates are otherwise well specified and executable."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The decision tree is dense but each path resolves: Step 4 fixes the QA Read order and installation fallback, Step 4.7 mandates smoke probes even for small diffs without plan/server, test_stub findings are forced to ASK with regression-before-repair in 5d, incomplete/blocked/not-run probes propagate to COMPLETED=false, and 5.8 defines reruns of affected probes with new REVIEW_START tokens. Residual ambiguity: Step 4.7.3 says 'post-expiry smoke rechecks are not-run' while also requiring post-expiry 'revalidation' and 5.8 requires rerunning affected smoke probes after fixes—the agent must infer that fix-triggered reruns count as revalidation, not rechecks. Step 4's 'research runs alongside specialist dispatch' also slightly muddles ordering with Step 4.5. Outputs, records and tool invocations are otherwise well specified."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "review/SKILL.md workflow",
|
||
"bundle": "negative",
|
||
"verdict": "fail",
|
||
"mean": {
|
||
"clarity": 2,
|
||
"completeness": 1,
|
||
"actionability": 1
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 1,
|
||
"actionability": 1,
|
||
"reasoning": "The bundle contains only section headings; no instruction bodies survive, so every step requires guessing. Example: review/SKILL.md 'Step 4.7: Exploratory QA (before Fix-First)' and qa/sections/exploratory.md '2. Probe loop'/'3. Parent handoff' define no charter, probe bound, incomplete-probe handling, or handoff format, so the agent cannot know what to run or report. Example: SKILL.md Step 5c/5d and '1. Re-review after edits' give no test_stub ASK gate rule or which evidence to rerun after approved repairs; review/sections/shared-code-reuse.md is entirely empty. Ordering is also inconsistent (Step 0 precedes the title; Steps 4.5/4.8 live in sections with no sequencing guidance). No deliverable schema is present beyond template headings."
|
||
},
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 1,
|
||
"actionability": 1,
|
||
"reasoning": "The bundle contains only headings; no step bodies, decisions, or outputs survive, so the agent must guess every action. Example 1: review/SKILL.md 'Step 4.7: Exploratory QA (before Fix-First)' and qa/sections/exploratory.md '2. Probe loop'/'3. Parent handoff' define nothing about running without a plan or server, what counts as an incomplete probe, or what the parent receives — the report-only discovery rule and test_stub ASK gate appear nowhere. Example 2: 'Step 5.8: Persist Eng Review result' → '1. Re-review after edits' has no instruction to rerun affected QA evidence after approved repairs, and review/sections/shared-code-reuse.md is empty despite being indexed. The skeleton is navigable, but nothing is executable."
|
||
},
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 1,
|
||
"actionability": 1,
|
||
"reasoning": "The bundle contains only headings; no step bodies, decisions, or outputs are present, so execution requires guessing everywhere. Example 1: review/SKILL.md places 'Step 0: Detect platform and base branch' before the title and section index, and Step 4.7 Exploratory QA, Step 4.5/4.6 Review Army, and Step 4.8 Adversarial sit across files with no visible Read triggers or ordering rule, leaving the agent unsure which pass runs when. Example 2: the stated test_stub ASK gate, report-only discovery rule, incomplete-probe honesty handling, and 'Re-review after edits' rerun-evidence requirement are named in headings only (Step 5c, Step 5.8 §1, exploratory.md §2) with no defined conditions or deliverables. shared-code-reuse.md is empty."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "review/SKILL.md workflow",
|
||
"bundle": "defect",
|
||
"verdict": "pass",
|
||
"mean": {
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "Paths are explicit and cross-referenced: small diffs still run Step 4.7 smoke (review-army.md small-diff branch), test_stub forces ASK (Step 5a), incomplete probes map to blocked/not-run and COMPLETED:false (Step 5.8), and reruns after approved repairs are defined (Step 5.8.2, Step 4.7.4). Defects are minor: Step 5.0 says save actions in the 'Step 3 action list' but no Step 3 exists (only 3.4–3.6; elsewhere called 'invocation action list'), and Step 5.8 repeats 'Steps 3–5' similarly. Also 'not applicable' QA outcomes (exploratory.md §4) aren't mapped onto Step 5.8's COMPLETED rule, leaving nonbehavioral-diff completion slightly underspecified. Otherwise executable without guessing."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "Dense but coherent decision tree: paths for small diffs, missing plan/server, blocked probes, test_stub ASK gates, rerun-after-repair and nonconvergence are each explicitly resolved, and section Reads are ordered. Minor defects: SKILL.md Step 5.0 'Keep decisions through fix cycles' tells the agent to save actions in 'the Step 3 action list', which is never defined (elsewhere called the 'invocation action list'); and Step 5.8/shared-code-reuse depend on a captured REVIEW_START token whose creation never appears in the supplied steps, only `--finish REVIEW_START`. Step 4's 'Research runs alongside specialist dispatch' also blurs read-only Step 4 with Step 4.5 timing. None force guessing on a permitted action, but they cost locating effort."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The decision tree is dense but internally consistent: Step 4 reads scope/methods read-only, Step 4.7 runs bounded smoke plus required plan checks (5 min/12 probes, no DEADLINE reset, post-expiry rechecks not-run), QA defects return with test_stub forced to ASK (Step 5a), repairs require red-then-green regression plus original/adjacent probes (5d), and Step 5.8 reruns affected probes and marks COMPLETED false on any blocked/not-run required probe. Minor defects: Step 5.0 'Keep decisions' tells the agent to save actions in a 'Step 3 action list' though no Step 3 exists (only 3.4–3.6), and Step 5.8 says 'a pass covers Steps 3–5' with similar numbering drift; REVIEW_START capture is assumed from the preamble. Neither forces a wrong action, but they cost lookup effort."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "setup-deploy/SKILL.md platform setup",
|
||
"bundle": "positive",
|
||
"verdict": "pass",
|
||
"mean": {
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "Steps 2–6 are concrete: detection script, per-platform branches, a field-collection rule before writing, a fixed CLAUDE.md template with explicit duplication rule, and non-blocking verification. Multi-platform detection and non-deploying projects are handled. Gaps: Fly computes the health check from the inferred `{app}.fly.dev` URL before asking the user to confirm a possibly custom domain, with no instruction to recompute, so the written health check can target the wrong host. Step 3 routes Heroku/Railway to the Custom questions (which ask the deploy trigger), yet Step 4 says Heroku/Railway leave the trigger unset, creating redundant or contradictory asks. Render checks RENDER_API_KEY but never uses the result, and the summary omits project type and hooks."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "Steps 2–6 form a coherent, executable pipeline: detection script, per-platform branches, explicit confirmation gating, a fully specified CLAUDE.md template with dual-field rules, verification, and summary. Decision points (multiple platforms, non-deploying projects, merge-method allowed via gh) are handled. Two gaps: Step 4 claims Heroku/Railway leave the deploy trigger unset, yet their branch routes through the Custom questions that explicitly collect the trigger—a small contradiction about what counts as confirmed. Vercel's 'production URL from vercel project settings' gives no command to obtain it, so the agent must fall back to asking without clear guidance. Otherwise unambiguous and deliverable-ready."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 4,
|
||
"actionability": 4,
|
||
"reasoning": "The supplied Steps 2-6 form a coherent, executable flow: a concrete detection script, per-platform sub-procedures with explicit confirmation gates, a CLAUDE.md template with a consistency rule for duplicated fields, non-blocking verification, and a summary. Multi-platform and project-type disambiguation are handled. Gaps remain: Step 4's field list omits `Deploy workflow`, so when fly.toml coexists with a GitHub Actions deploy workflow (or several workflows match), it is undefined which file is recorded or how to choose among them. Step 4 also says to ask for any missing status command while the template offers an 'HTTP health check' fallback, leaving Vercel/Netlify agents unsure whether to prompt or default. These are minor ambiguities in an otherwise well-specified workflow."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "setup-deploy/SKILL.md platform setup",
|
||
"bundle": "negative",
|
||
"verdict": "fail",
|
||
"mean": {
|
||
"clarity": 1.3333333333333333,
|
||
"completeness": 1,
|
||
"actionability": 1
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 2,
|
||
"completeness": 1,
|
||
"actionability": 1,
|
||
"reasoning": "The supplied workflow is only a skeleton of headings with no instructions beneath them. Step 2 'Detect platform' lists 'Platform config files' and 'GitHub Actions deploy workflows' but gives no file names, glob patterns, or precedence rules when multiple platforms are found. Step 4 'Write configuration' names a CLAUDE.md section header but specifies no fields (URLs, status commands, health checks) or format to persist. Step 1 is absent, and Steps 5-6 have no verification commands or summary content. The outline is readable, but an agent cannot detect a platform, gather config, or produce a CLAUDE.md deliverable from headings alone."
|
||
},
|
||
{
|
||
"clarity": 1,
|
||
"completeness": 1,
|
||
"actionability": 1,
|
||
"reasoning": "The supplied workflow consists solely of section headings with no instructional body. For example, 'Step 2: Detect platform' lists 'Platform config files' and 'GitHub Actions deploy workflows' but gives no file names, glob patterns, or detection logic to distinguish Fly.io from Render or Vercel. Likewise 'Step 4: Write configuration' names a CLAUDE.md heading and 'Custom deploy hooks' but specifies no fields (URLs, status commands, health checks) or format to persist. Step 1 is absent entirely, and Steps 5–6 have no verification criteria or summary content. An agent cannot execute any step or produce the expected CLAUDE.md deliverable from this outline alone."
|
||
},
|
||
{
|
||
"clarity": 1,
|
||
"completeness": 1,
|
||
"actionability": 1,
|
||
"reasoning": "The supplied entrypoint is only a heading skeleton with no instruction body. For example, '### Step 2: Detect platform' lists '# Platform config files' and '# GitHub Actions deploy workflows' but gives no file names (fly.toml, render.yaml, vercel.json) or detection logic, so an agent cannot determine which platform applies. Likewise '### Step 4: Write configuration' names a '## Deploy Configuration' section and 'Custom deploy hooks' but provides no schema, fields, or prompts for URLs, status commands, or health checks. Step 1 is absent entirely, and Steps 5-6 have no verification criteria or summary format. Nothing here is executable; the agent would have to invent the whole workflow."
|
||
}
|
||
]
|
||
},
|
||
{
|
||
"judge": "setup-deploy/SKILL.md platform setup",
|
||
"bundle": "defect",
|
||
"verdict": "fail",
|
||
"mean": {
|
||
"clarity": 4,
|
||
"completeness": 3,
|
||
"actionability": 3.3333333333333335
|
||
},
|
||
"samples": [
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 3,
|
||
"actionability": 4,
|
||
"reasoning": "Platform sections are concrete and executable (Fly extraction/health-probe logic, Render key check that avoids leaking bytes), and Step 4 explicitly resolves which fields each path leaves unset, lists allowed merge methods via gh, and defines the no-deploy case. Two gaps weaken it: Step 2 says 'Run the platform detection from the deploy bootstrap:' followed by an empty block, so the agent has no command or expected output format to branch on in Step 3. Also Step 4 claims Heroku/Railway leave the deploy trigger unset, yet that path routes through the Custom/Manual questions which ask for the trigger; and the template defaults differ ('HTTP health check' vs 'poll production URL') despite the instruction that both fields hold identical values."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 3,
|
||
"actionability": 3,
|
||
"reasoning": "Platform-specific branches are precise (exact grep/sed, curl, and gh commands; explicit confirmation rules; well-defined CLAUDE.md template with duplicated hook fields explained). But Step 2 is hollow: 'Run the platform detection from the deploy bootstrap:' is followed by nothing, so the agent has no command or file reference to actually detect Fly/Render/Vercel/etc., undermining every downstream branch. Additionally, the Vercel path tells the agent to use 'the production URL from vercel project settings' without a command to retrieve it, and Heroku/Railway detection is routed to a section gated on 'If nothing detected', a minor ordering contradiction. Steps 4–6 are otherwise executable and produce the deliverable."
|
||
},
|
||
{
|
||
"clarity": 4,
|
||
"completeness": 3,
|
||
"actionability": 3,
|
||
"reasoning": "Platform sections, the CLAUDE.md template, and verify/summary steps are concrete and mostly unambiguous, with explicit handling for multi-platform detection, non-deploying projects, and secret-safe key checks. The decisive gap is Step 2: 'Run the platform detection from the deploy bootstrap:' is followed by nothing, so the agent has no command or criteria for how fly.toml/render.yaml/workflows are actually detected, undermining everything downstream. Secondary looseness: Vercel's health check is 'the production URL from vercel project settings' with no command given, and Render sets the health check to an inferred URL before user confirmation. The workflow is executable once detection is supplied, but as written it depends on an undefined step."
|
||
}
|
||
]
|
||
}
|
||
]
|
||
}
|