Commit Graph
1098 Commits
Author SHA1 Message Date
Garry Tanandcapy-ai[bot] 2714e9f7f3 READY prints real trial merges and the second-model trail read, plus six small ship-tool fixes (GST-3200) (#4529)
READY now prints each open PR's trial-merge result and the reader of the
second-model trail read, or a loud MISSING line, instead of lines left
to fill in.

Why: agents dropped those lines as noise and skipped the trail read on
three gate-class PRs. Six more small ship-tool bugs from 2026-11-19
reports are fixed here too.

Nothing to do: pass `--trail ` to `gstack-ship-queue ready` on
risk-class PRs.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/outside-voice#1 reader=outside-voice
Review-Finding: p1/outside-voice#2 reader=outside-voice
Review-Finding: p1/outside-voice#3 reader=outside-voice
Review-Finding: p1/outside-voice#4 reader=outside-voice
Review-Finding: p1/outside-voice#5 reader=outside-voice
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/adversarial-native#4 reader=adversarial-native
2026-11-19 17:05:16 +00:00
50cc282273 Test runs whose workers start in their own session are no longer killed as stalled (GST-3213) (#4533)
/ship's test check no longer kills a quiet, working test run as stalled
when the test command starts its workers in their own session, as pnpm
11's `pnpm exec` does for Vitest.

Why: the stall check read CPU time from the command's process group
only, where just the idle launcher was left; it now counts every process
descended from the command.

Nothing to do: a truly hung run is still killed at the same stall
timeout, and every threshold is unchanged.

Co-authored-by: dtoelke <14206719+dtoelke@users.noreply.github.com>
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 16:54:44 +00:00
104fc72757 Syncing a second repo no longer makes GBrain ask the first repo to sync again (GST-3212) (#4532)
Once you synced a second repo, every repo synced before it showed a
false "run /sync-gbrain" prompt at session start. The read check now
uses the per-repo sync record `/sync-gbrain` already keeps.

Why: people who work across several repos saw this on every session.
Nothing to do after upgrading.

Co-authored-by: SatriaSuriaAriffin <57977752+SatriaSuriaAriffin@users.noreply.github.com>
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 16:42:32 +00:00
Garry Tanandcapy-ai[bot] 9610ba43db gstack's Codex hooks keep working with unusual folder paths and after uninstall (GST-3197) (#4528)
gstack's Codex hooks no longer block every prompt when your Codex
folder's path has a `(` or `&`, and uninstall keeps your own Codex hooks
trusted.

Why: both broke Codex sessions; setup also now matches Codex's hook
rules, and the /review checklist on Codex reads shell commands more
carefully.

Nothing to do: plain paths keep the exact hook command they had, so
Codex asks for no new trust.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=red-team
Review-Finding: lib/codex-rollout.ts:217:functional-contract reader=qa
2026-11-19 15:42:21 +00:00
Garry Tanandcapy-ai[bot] bd449234aa Pre-commit /verify fails on a timeout, and the review collector stops reading reviewers' scope lists as findings (GST-3201) (#4527)
Pre-commit /verify now fails a check that times out and runs its checks
on one time budget; a reviewer's scope-list bullets no longer need an
owner ruling to clear; stack tests fail when git can't list a PR's
tests.

Why: each let a result read clean with nothing proving it, or block on
nothing (#4518, #4519). Goal: gates never pass on missing evidence.

Nothing to do; `gstack-checks quick --timeout` is now one budget the
checks share (540 s default).

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/outside-voice#1 reader=outside-voice
Review-Finding: p1/outside-voice#2 reader=outside-voice
2026-11-19 15:17:30 +00:00
Garry Tanandcapy-ai[bot] 3cccf074ad Free-suite CI survives a slow Ubuntu mirror, and four flaky tests and eval checks stop failing (GST-3206) (#4530)
Free-suite CI stops failing before any test runs when archive.ubuntu.com
is slow, and four tests that failed and then passed in the last day are
fixed at their cause.

Why: 27 setup failures and 5 flaky reds on 2026-11-19 cost reruns and
blocked READY checks across PRs and main.

Nothing to do: the first run on each runner image fills the new package
cache.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 15:16:23 +00:00
Garry Tanandcapy-ai[bot] 17303862d0 Destructive-command check closes the parsing gaps found after #4513 and blocks wipefs -a in every spelling (GST-3190) (#4525)
The destructive-command check now blocks force-pushes to main written
with `--repo=`, after a `cd`, or with `--all`, and asks again for the
other shapes that slipped past it after #4513.

`wipefs -a` is blocked in every spelling, including targets fed by
`xargs`; only existing image files it names outside `/dev` are exempt.
Dry runs and greps for SQL text stop over-blocking.

Nothing to do: the check runs in every session already.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 14:00:44 +00:00
Garry Tanandcapy-ai[bot] 18e24a2ee9 Codex /review with gpt-6.1-sol counts the second-model pass it already ran (GEVA-562) (#4526)
On Codex with gpt-6.1-sol, /review is no longer blocked at session end
for a missing second-model pass that it ran.

Why: with Sol, Codex starts subagents through its `collaboration` tools
and names a started agent by `task_name`, not `agent_id`, so the
checklist reader missed it; this blocks gstack-evals' Codex pin.

Nothing to do: the fix ships with the next gstack release.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 13:55:28 +00:00
Garry Tanandcapy-ai[bot] 45e6d549e3 A second review round reviews a gate-class fix instead of reporting done (GST-3187) (#4520)
/ship's second review round now reviews a fix that touched a gate-class
path, instead of reporting done with nothing read.

A plan told "nothing was fixed" while the branch has new commits since
the reviewed head now refuses with `review_required_no_readers` rather
than printing done.

Nothing to do: the next /ship picks it up.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/red-team#1 reader=red-team
2026-11-19 13:33:04 +00:00
Garry Tanandcapy-ai[bot] b23e1b01a7 Codex picker prompts and docs name skills the way Codex lists them (GST-3196) (#4524)
On Codex, the suggested prompts for nine gstack skills and the README
now say `$office-hours`, `$review` and so on, which load the skill;
`$gstack-office-hours` loaded nothing.

Why: Codex lists each skill by the `name:` in its SKILL.md, not the
`gstack-` folder, and skill_prefix never changes that name on Codex.

Nothing to do: the next `./setup --host codex` picks it up.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 13:10:18 +00:00
Garry Tanandcapy-ai[bot] b0f865c00c Codex now holds /review to its checklist before a session can end (GST-3173) (#4522)
On Codex, gstack's /review can no longer end a session with a required
step skipped: the end-of-session check reads Codex's own session log and
sends the agent back to finish, as it already does on Claude Code.

Why: Codex users got routing and the destructive-command check in #4512,
but not the /review checklist, so a review could stop after the diff
read with no second-model pass.

To use it: run `./setup` again and trust the new hook in Codex
(`/hooks`); turn it off with `GSTACK_SKILL_CONTRACTS=off`.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=testing
Review-Finding: p1/specialist-batched#2 reader=testing
2026-11-19 12:44:06 +00:00
Garry Tanandcapy-ai[bot] be83949eb6 Codex skill suggestions name each skill the way Codex lists it (GST-3195) (#4523)
On Codex, gstack's routing hint now names each skill by the `name:` in
its installed SKILL.md (`review`), the name Codex lists, so `$review`
loads the suggested skill.

Before, the hint said `gstack-review` (the folder name), which Codex
doesn't list, so mentioning it loaded nothing.

Nothing to do: the next `./setup --host codex` picks it up. Claude
Code's hint is unchanged.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 11:59:36 +00:00
Garry Tanandcapy-ai[bot] f82b6f76f3 A failed second-model review names the cause that stopped it, and never deletes its own prompt (GST-3188) (#4519)
When a second-model review fails, gstack now reports the failure that
actually ended it: Codex's final error decides, a stray "402" in a
request id no longer reads as a used-up budget, and a Claude Code 402 or
usage limit reads as a budget stop with no Codex advice.

Why: these were the four follow-ups ruled below the blocking bar on
#4515; each one could send someone to fix the wrong thing.

Nothing to do. The prose rule that reads "auth", "login" or "API key" in
stderr as a login failure is not changed here (see below).

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 11:50:17 +00:00
Garry Tanandcapy-ai[bot] e95cea8404 On Capy, /ship hands its pull request edits and comments to Capy's PR tool, always naming the right repository (GST-3183) (#4518)
On Capy, /ship's PR body edits, comments and new PRs no longer go out
through `gh` under the machine owner's GitHub account.

Each write runs its checks, saves the exact text to a private file and
prints a `HOST_PUBLISH:` line that names the repository (and host, off
github.com) it belongs to, so Capy's PR tool sends it to the same place
`gh` would have.

Nothing changes outside Capy (`GSTACK_CAPY` unset).

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=specialist-batched
Review-Finding: p1/specialist-batched#3 reader=specialist-batched
Review-Finding: p1/red-team#6 reader=red-team
Review-Finding: p1/specialist-batched#1 reader=testing
Review-Finding: p1/specialist-batched#2 reader=testing
2026-11-19 11:30:53 +00:00
Garry Tanandcapy-ai[bot] ed267248d3 A folder marked as published data can no longer hide source code or agent files from /ship's review (GST-3158) (#4516)
In a folder a repo marks as published data (`review_content_patterns`),
source code in any language, agent instruction files, config and data
that kept code names by full path now stay in /ship's review.

Why: #4511 let a `.java` file, a binary `AGENTS.md` or a file named by a
kept script leave review there, and reviewers must see every line of
code a PR changes.

Nothing to do: no base policy sets the key yet.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/specialist-batched#1 reader=specialist-batched
2026-11-19 11:15:51 +00:00
Garry Tanandcapy-ai[bot] 944b41e24d Codex gets gstack's skill routing hint and destructive-command check (GST-3172) (#4512)
`./setup --host codex` now adds gstack's skill-routing hint and
destructive-command check to Codex's `~/.codex/hooks.json`, next to your
own hooks, so Codex users get the same guardrails Claude Code users
have.

Codex runs a hook only after you trust it, so nothing fires until you
accept gstack's hooks at Codex's startup prompt or in `/hooks`;
`gstack-doctor` shows whether you have.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/red-team#1 reader=red-team
2026-11-19 11:09:11 +00:00
Garry Tanandcapy-ai[bot] 55aaa6c4a5 Destructive-command check stops asking about scratch deletes and blocks real damage (GST-3169) (#4513)
The destructive-command check no longer asks before deleting temp, cache
or build folders, and now catches dangerous commands it used to miss:
deletes outside the project, force-pushes to main inside longer
commands, history rewrites, database resets and cloud deletes.

Agents were stopped for harmless cleanup while real damage got through:
on a 102-row replay of real command shapes, main gives the expected
answer on 66 and this branch on all 102.

Nothing to do, except: an Agent SDK app where a person answers prompts
should set `GSTACK_SESSION_KIND=interactive`.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=security
Review-Finding: p1/specialist-batched#2 reader=security
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/red-team#2 reader=red-team
Review-Finding: p1/red-team#3 reader=red-team
Review-Finding: p1/red-team#4 reader=red-team
Review-Finding: p1/red-team#5 reader=red-team
Review-Finding: p2/red-team#1 reader=red-team
2026-11-19 10:40:49 +00:00
Garry Tanandcapy-ai[bot] 957fab79c5 Second-model review keeps the reviewer's result and reports budget stops (GST-3181) (#4515)
When Codex reviews a change, gstack now keeps the review file the
reviewer wrote instead of replacing it with Codex's one-line closing
message, and a provider that refuses with HTTP 402 is reported as a
used-up budget instead of "exit 1" or a login failure.

Why: gstack-evals lost 22 review attempts to the overwrite and 5 of 45
sessions to a 402 from a metering proxy whose cause gstack threw away.

Nothing to do: a 401 or 403 is still a login problem. Open notes, one of
them a known gap, are listed below.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=testing
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/red-team#2 reader=red-team
2026-11-19 10:27:37 +00:00
Garry Tanandcapy-ai[bot] f13dbb0d4a Ship policy can mark published data as content, so /ship can review large data PRs (GST-3158) (#4511)
A repo can now list published-data folders under a new
`review_content_patterns` ship-policy key, so /ship leaves `.patch`
diffs, `MANIFEST.sha256` and extension-less licences out of review,
while source files there stay reviewed.

Why: gstack-evals #352 is 13.8 MB, 13.6 MB of it published session data,
and could never pass /ship's 2 MB review limit.

Nothing to do until a repo adds the key to its base-branch policy; a
pattern that keeps too many files now stops with
`REVIEW_EXCLUDE_KEEP_TOO_LARGE` instead of being dropped.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/red-team#2 reader=red-team
Review-Finding: p2full/red-team#1 reader=red-team
2026-11-19 09:02:33 +00:00
Garry Tanandcapy-ai[bot] c25390daa2 /ship's review works when the shipped checkout isn't the project's default checkout (GST-3145) (#4507)
/ship's review pass now lets its readers start in a checkout of an
unrelated repository, and refuses to start them in any other checkout
that holds the review's diff base, such as a sibling worktree or a
separate clone of the shipped repository.

Why: a project whose default checkout is another repository needed a
fresh machine for every /ship review, while a worktree or clone of the
same history could read the wrong diff with no error.

Nothing to do: the pass still stops before anything starts if the
shipped checkout can't be reached.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 08:55:01 +00:00
Garry Tanandcapy-ai[bot] 12bd33b648 Pull requests that merged a release reuse their earlier test run when the release only regenerated a declared file (GST-3067) (#4483)
When a pull request merges a release whose only extra change is a
generated file the repo lists under `release_outputs`, its earlier
remote test run now counts for it, if the clear command gets
`--allow-repo-commands`.

Why: gbrain paid one extra full CI run for each such pull request,
several a day.

Without the flag nothing changes except that the refusal now says to
pass it; a hand edit to the generated file, or any other changed file,
is still refused.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-19 08:11:34 +00:00
Garry Tanandcapy-ai[bot] ed4086e262 Codex loads office-hours, plan-design-review and gstack-upgrade again, and knows when it runs unattended (GST-3161) (#4510)
On Codex, naming office-hours, plan-design-review or gstack-upgrade
loaded nothing, because setup left a second copy of each that Codex
listed twice; on macOS and Linux setup now links those copies, and
re-running setup or upgrading fixes an existing install.

Skills started under `codex exec` now report `SESSION_KIND: headless`
instead of `interactive` (also on slim Linux images without `ps`), Codex
reviewers start without the parent's conversation (`fork_turns:
"none"`), and the docs stop saying Codex has no hooks.

Nothing to do after merge; Codex users pick it up on their next upgrade.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/specialist-batched#1 reader=testing
2026-11-19 07:18:52 +00:00
Garry Tanandcapy-ai[bot] b702243859 When Claude Code runs /verify before a commit, it now gets a quick check instead of the full fix loop (GST-3150) (#4508)
When Claude Code calls /verify on its own right before `git commit`,
gstack now runs one quick check-only command that ends in a single
`VERIFY PASS`, `VERIFY FAIL` or `VERIFY skipped: ` line; typing /verify
still runs the full skill.

Why: since Claude Code 2.1.286 any skill named `verify` is run before
every commit, so gstack's fix loop, its first-run notices and a hard
stop in repos with no remote landed on ordinary commits.

Nothing to do; `gstack-config set verify_before_commit off` turns the
check off.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=specialist-batched
Review-Finding: p1/specialist-batched#2 reader=specialist-batched
2026-11-19 06:06:43 +00:00
Garry Tanandcapy-ai[bot] 5afef04e31 The prompt-routing hook suggests skills you install, and stops sending bug reports to /qa (GST-3123) (#4506)
The prompt-routing hook can now suggest skills you installed yourself:
any skill that lists `triggers:` in its SKILL.md, not only gstack's
/review, /qa and /verify. It also stops sending ordinary bug reports to
/qa.

Why: a bug-fix skill you install was never named (0 of 60 real bug
reports); now it's named on 40 of 60, and /qa takes none of them instead
of 3.

Nothing to do after merge.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/specialist-batched#2 reader=api-contract
Review-Finding: p1/specialist-batched#1 reader=security
2026-11-19 04:26:51 +00:00
Garry Tanandcapy-ai[bot] 3f1b82b22a Helper hooks are on by default, the destructive-command check runs as a hook, and prompt routing fires on long bug reports (GST-3110) (#4503)
Claude Code now gets gstack's prompt-routing, skill-contract and
destructive-command hooks on by default, and new OpenClaw and Hermes
registrations get the start hook. On 60 real bug reports, routing now
matches a bug-fix skill on 40 (was 22), but gstack's own skills route 14
(was 17).

Why: on Opus 5.5, unattended, a hook line naming a skill got it loaded
in 58 of 60 sessions, and a skill relying on its description in 0 of 60
(gstack-evals#369).

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1v/outside-voice#1 reader=outside-voice
Review-Finding: p1v/adversarial-native#4 reader=adversarial-native
Review-Finding: p2/red-team#1 reader=red-team
2026-11-18 23:36:58 +00:00
Garry Tanandcapy-ai[bot] 43026970b7 /ship's state-machine check accepts the header lines its reader is told to write (GST-3115) (#4504)
The state-machine check in /ship no longer stops a review because its
reader opened the reply with the `REVIEWED_TREE` and `REVIEW_VERDICT`
lines its instructions require.

Why: on yc-software/qm#3728 the check reported those two lines as gaps,
so /ship stopped before its adversarial review step.

Nothing to do after merge.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 22:33:08 +00:00
Garry Tanandcapy-ai[bot] 2f57b07495 gstack's script runner and merge-queue checks now refuse cut-off command output (GST-3104) (#4502)
The helper that runs gstack's own scripts and `bash -c` commands, and
the merge queue's branch-ownership, stacked-test, release-version and
merge-group record checks, now fail with `SPAWN_OUTPUT_INCOMPLETE` when
Bun cuts a command's output off, instead of treating part of it as the
whole answer.

Why: #4492 added this for the scanners and gates but left these on
Node's `spawnSync`, and its release note claimed wider coverage than it
shipped.

Nothing to do: a new test stops any new gstack command from calling
Node's `spawnSync` directly.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 18:40:12 +00:00
d789dd965b /sync-gbrain's read check works when the first page is over 64 KB (#4396) (#4497)
/sync-gbrain's read check no longer reports `unknown` when the first
page in your code source is over 64 KB.

Why: the check read the whole page to confirm its source and slug, and
big pages (lockfiles) hit the 64 KiB output limit on every run.

Nothing to do: it now reads a 4 KB window of the page and still treats
any cut-off output as `unknown`.

Co-authored-by: nfitzsimmons <2198211+nfitzsimmons@users.noreply.github.com>
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=testing
2026-11-18 17:08:51 +00:00
0ab5657112 Changed-number check stops flagging tests that only share a word like limit (#4478) (#4496)
/ship's changed-number check no longer stops on tests that only use the
same word as a changed property: lowering `limit: 120` to `30` used to
report every test that said "limit" and held 120.

Why: `docs/pregate.md` already said a test counts for a word like
`limit` only when it also names the changed module, but the code
enforced that only for one-character values. Tests that import the
module are still reported.

Nothing to do. Thanks to @dtoelke, who reported this and sent the first
fix in #4478.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Co-authored-by: dtoelke <14206719+dtoelke@users.noreply.github.com>
2026-11-18 17:03:55 +00:00
Garry Tanandcapy-ai[bot] 36b349d660 Commands that read cut-off output now fail instead of judging part of it (GST-3086) (#4492)
Commands that read large git or gh output, including the credential
pre-push hook, the pre-merge secret and literal scans and the
sealed-output guard, now fail when Bun cuts that output off, instead of
judging the part they got.

Why: Bun can report exit status 0 together with ENOBUFS and partial
stdout, so a push whose diff passed 64 MiB could get a credential past
the pre-push hook.

Nothing to do: the failure names `SPAWN_OUTPUT_INCOMPLETE` and links its
troubleshooting entry.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 17:00:18 +00:00
Garry Tanandcapy-ai[bot] 2767d229fb gstack-artifact validate now refuses a run.json that gstack-autoplan never wrote (GST-3090) (#4494)
`gstack-artifact validate` and `ack` now refuse a run.json that
gstack-autoplan never wrote, so a hand-built run can no longer pass as a
finished review.

Why: a session skipped `gstack-autoplan next`, built run.json with a
script copied from the reference fixture, and got ARTIFACT_VALID.

Nothing to do unless you validate old or hand-made run directories:
rerun those through gstack-autoplan.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 16:45:06 +00:00
Garry Tanandcapy-ai[bot] 45f879b534 The Codex readiness check and the second-model review pick the same model for a custom Codex provider (GST-3089) (#4491)
With a custom Codex provider, `gstack-codex-probe role-ready` and
`gstack-outside-voice run --runner codex-cli` now resolve the
plan-review model the same way, so a session no longer skips a review
that would have run.

Why: the check refused a `config.toml` naming `gpt-6-astra` while the
runner ran it, and a day-one eval session dropped the second-model
review in every phase.

Nothing to do: a custom provider naming the frontier catalog ID just
works; any other name now says how to opt in.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 16:44:07 +00:00
Garry Tanandcapy-ai[bot] 3610130d6e Keep a policy-floor test passing after gstack's version bump to 1.101 (#4495)
A policy-floor test assumed gstack's own version stays below the next
minor after the newest policy key. The v1.101 release (#4493) reaches
that number, so its CI fails; the test now checks a floor above both.

Why: this unblocks today's stable release, which carries a security fix
(#4486) and four blocker fixes.

Nothing to do as a reader.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 16:25:47 +00:00
Garry Tanandcapy-ai[bot] fda9dea79f /ship on Codex starts its audits with Codex's own subagent tools, not Claude Code's (#4476) (#4489)
On Codex, /ship and the other skills now start audit and reviewer
subagents with Codex's own `spawn_agent` and wait on `wait_agent`,
instead of Claude Code's Agent tool, which Codex can't call.

Why: Codex's /ship stopped at the Step 7 coverage audit and never
finished (#4476).

Codex users need `multi_agent = true` under `[features]` in
`~/.codex/config.toml`.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 15:48:59 +00:00
Garry Tanandcapy-ai[bot] 782f4595ed Two owner-answer adds at once no longer collide, and three flaky tests get real fixes (GST-3084) (#4490)
Recording the same owner answer from two processes at once no longer
fails about one time in ten; the second add waits for the first to
finish writing.

Why: this was a daily flaky-test repair, and the owner-answers "two
processes adding at once" failure turned out to be a real race, not a
flaky test.

Nothing to do: tests only change, except the 11 lines in
`lib/owner-answers.ts`.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=testing
2026-11-18 15:08:17 +00:00
Garry Tanandcapy-ai[bot] bf2c126b55 gstack-post no longer reads /ship's own run ids and commit ids as phone numbers (#4480) (#4488)
Posting a /ship PR body no longer stops on a phone-number warning for
the run id and commit ids that /ship writes into it.

Why: most /ship runs hit this confirm (#4480, and #4474), which turned a
privacy guard into noise people click past.

Nothing to do: the next /ship post goes through without `--confirm`.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: v1/outside-voice#1 reader=outside-voice
Review-Finding: v1/adversarial-native#1 reader=adversarial-native
2026-11-18 14:57:04 +00:00
Garry Tanandcapy-ai[bot] 3ecd7168b2 The token-capped /ship child on Claude Code now starts its prompt with the typed /ship command (GST-3078) (#4487)
The /ship child that `gstack-ship-budget launch` starts on Claude Code
now opens its prompt with `/ship --continue ` (or `/gstack-ship` under
`skill_prefix`), so Claude Code expands the skill instead of hoping the
model finds it.

Why: Claude Code expands a skill only when the prompt starts with the
typed command, and unattended Opus 5.5 found a skill from its
description in 0 of 60 sessions (gstack-evals#366).

Nothing to do: the rule for parents is now in docs/unattended.md.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 13:32:03 +00:00
Garry Tanandcapy-ai[bot] 0109703797 Security findings no longer show their file, line or reason in PR bodies (GST-3076) (#4486)
A security finding that an owner ruled a note, dispositioned, or that
review deferred now shows in a PR body only as an opaque id and
"security finding, details recorded privately".

Why: on a public repository the old lines named the file, line and
reason of a live exposure outside the PR's change.

Nothing to do: other findings print exactly as before, and the details
stay in the private approval record.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=security
Review-Finding: p1/red-team#1 reader=red-team
2026-11-18 12:57:24 +00:00
Garry Tanandcapy-ai[bot] 07611a00a7 /ship no longer reports zero blockers when the second-model review found one (GST-3075) (#4485)
When /ship stopped on a second-model review finding marked "Blocks:
yes", the record it left on the PR said no blockers were found; now the
finding is counted and /ship will not write that record until it is
fixed or ruled.

Why: `ready` and the merge step trust that record, so an undercount let
a blocked PR look clean.

Nothing to do: the next /ship run logs Step 11's findings and checks
them before writing.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: h2p1/red-team#1 reader=red-team
Review-Finding: h2p1/red-team#2 reader=red-team
Review-Finding: h2p1/red-team#3 reader=red-team
Review-Finding: v1/outside-voice#1 reader=outside-voice
2026-11-18 12:51:14 +00:00
Garry Tanandcapy-ai[bot] bbf48aa7ef Unattended /autoplan records the owner's relayed gate reply instead of refusing it (GST-3068) (#4484)
An unattended /autoplan run that's handed the owner's reply at the final
gate now records it and finishes, the same way whichever instruction
file the agent reads.

Before, autoplan/SKILL.md said "never approve" while autoplan/DRIVER.md
said to record the relayed reply, so the outcome depended on which file
the model followed.

Nothing to do; this fixes existing instructions and adds no new feature.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 11:29:02 +00:00
Garry Tanandcapy-ai[bot] 44170d6791 Unattended /autoplan on Claude Code says its phase guard is on, and a report hidden in reasoning gets a clear fix (GST-3062) (#4481)
An unattended /autoplan run on Claude Code now says its phase guard is
on, instead of saying it is off while the guard blocks it. A phase
report written only in reasoning now gets a denial that says so and
names the exact command to send.

Why: in the day-one eval, two runs were blocked by a guard their own
record called off, and one run ended blocked after four publish attempts
whose report was invisible.

Nothing to do.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 08:47:07 +00:00
Garry Tanandcapy-ai[bot] f87d4325a3 A round grant at the review round cap buys the fix round it names (GST-2968) (#4479)
When /ship stops at a pull request's review round cap, a grant naming
that stop now buys one scoped review of the fix, in the same run or the
next, instead of another read-only full review.

Before, a grant recorded after the read-only review, or between /ship
runs, bought nothing, so teams paid for extra full reviews or second
grants (ghome#40: $8.78 against about $3).

Nothing to do: /ship prints the grant command when it stops.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=security
Review-Finding: p1/specialist-batched#2 reader=state-machines
Review-Finding: p1/specialist-batched#3 reader=testing
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/red-team#2 reader=red-team
Review-Finding: p1/red-team#3 reader=red-team
2026-11-18 08:34:39 +00:00
Garry Tanandcapy-ai[bot] 33b1a965ba /ship's push step ends with one line that says whether it pushed (GST-3053) (#4474)
/ship's push step now always ends with one line that says what happened:
`PUSH: pushed …`, `PUSH: already_pushed …` or `PUSH: failed reason=`,
with the remote, branch and full commit.

Why: an eval grader looked for a `git push` call and counted 48 real
/ship pushes as "no push"; scripts and graders can now match this one
line instead.

If a script of yours matched the old `PUSH: ALREADY_PUSHED ` or `PUSH:
pushed to origin/` lines, update it; the exit status is unchanged.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 03:56:41 +00:00
Garry Tanandcapy-ai[bot] 9222485040 Review planner sends Swift health, permission and on-device storage code to full review (GST-3052) (#4472)
/ship's review planner now sends Swift code that asks for a privacy
permission, reads HealthKit or calendar data, or writes, deletes or
protects users' files on the device to the full review panel.

Why: since #4469, iOS changes that asked for microphone permission,
wrote and deleted meeting transcript files, or read calendars got the
light, one-reader review. Reported from the GHome iOS app's PR review.

Nothing to do: plain SwiftUI views, UserDefaults settings and file reads
still plan light.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/red-team#2 reader=red-team
2026-11-18 03:55:13 +00:00
Garry Tanandcapy-ai[bot] 0a6bb42e74 Reused CI runs that passed only after a re-run are now marked instead of shown as clean passes (GST-3054) (#4473)
When /ship reuses a CI run that passed only after its failed jobs were
re-run, its CI line and its final summary line now add `rerun=
attempts=`.

Why: gbrain's CI now records which test files passed on a re-run
(gbrain#9690), and a pass on a re-run must read as flaky, never as
clean.

Nothing to do: runs without that record print as before; a malformed
record is refused with `CI_CARRY_RERUN_MALFORMED`.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/outside-voice#1 reader=outside-voice
Review-Finding: p1/outside-voice#2 reader=outside-voice
Review-Finding: p1/adversarial-native#2 reader=adversarial-native
Review-Finding: p1/red-team#1 reader=red-team
2026-11-18 03:47:25 +00:00
Garry Tanandcapy-ai[bot] 237371474b Let autoplan revise accept a plan without section headings, as next does (GST-3050) (#4470)
`gstack-autoplan revise` now accepts a revised plan whenever the first
`next` accepted its shape, so a run no longer pays for a full review
round and then refuses the owner's edit.

Why: a plan with no section headings is reviewed whole by `next`, but
once the owner added one heading, `revise` stopped with
`PLAN_SECTIONS_INVALID`.

Nothing to do: runs already in progress pick this up too.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 02:21:31 +00:00
Garry Tanandcapy-ai[bot] c9b6fe0ab3 Second-model plan reviews in autoplan stop being reported unavailable; --help stops printing a fake refusal (GST-3051) (#4471)
A second-model plan review in `/autoplan` is no longer reported as
unavailable when it actually ran, and `gstack-outside-voice --help` no
longer prints a line that reads like a real consent refusal.

Why: sessions believed the false label and paid to re-run the review, or
closed the plan phase without its second-model review.

Nothing to do: the command autoplan already runs now picks the right
check by itself.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2026-11-18 02:12:43 +00:00
Garry Tanandcapy-ai[bot] 29e950227b Review planner sorts Swift and Xcode files by risk instead of sending every iOS PR to the full panel (GST-3041) (#4469)
/ship's review planner now sorts Swift and Xcode files by risk, so a
plain iOS UI change gets the light review instead of the full
five-reader panel.

Why: `gstack-diff-scope` had no rule for `.swift`, so every all-Swift
diff printed `SCOPE_ERROR=unmatched` and the planner failed closed.
Reported from the GHome iOS app's PR review.

Nothing to do: unknown file types still fail closed exactly as before.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/outside-voice#1 reader=outside-voice
Review-Finding: p2/red-team#1 reader=red-team
2026-11-18 01:28:22 +00:00
Garry Tanandcapy-ai[bot] dd4266dc4b Check each app build with an agent from another model family before its owner is asked to try it (GST-3025) (#4465)
Adds `gstack-first-use`: before an app build reaches its owner as "ready
to try", an agent from another model family plays that exact build, does
the plan's first five day-one actions with a screenshot or video of
each, and checks every link. Only then does it print the "try this
build" message.

Why: a recent app build passed every review and its owner still found
the first screen broken, because nothing tried the build the way the
owner would.

Nothing to do now; the first real use is that app's next build.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/red-team#2 reader=red-team
Review-Finding: p1/red-team#4 reader=red-team
2026-11-17 23:19:24 +00:00
Garry Tanandcapy-ai[bot] 0b656a171d Check a question against the owner's earlier answers before asking it again (GST-3025) (#4467)
An agent can now check a draft question against the owner's earlier
answers before sending it, so the owner isn't asked again about
something they already said yes to.

Why: owners repeat "don't ask me, just do it" across threads, because
each new session forgets earlier answers. In a four-day sample, about 1%
of 444 asks were repeats this could skip, so a two-week pilot decides
whether it's kept.

Nothing to do: the new command is opt-in, and no existing command,
install output or factory rule changes.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/specialist-batched#1 reader=testing
Review-Finding: p1/specialist-batched#2 reader=state-machines
Review-Finding: p1/red-team#1 reader=red-team
Review-Finding: p1/red-team#2 reader=red-team
Review-Finding: p3/outside-voice#1 reader=outside-voice
Review-Finding: p3/outside-voice#2 reader=outside-voice
2026-11-17 23:13:14 +00:00