Files
gstack/hosts/claude/hooks/question-preference-hook.ts
T
94993f7401 v1.61.0.0 fix wave: guards failing open / silent failures (9 fixes, 4 community PRs absorbed) (#2472)
* fix(careful): warn on chained rm even when the last target is safe

The safe-exception block whitelisted rm -rf of build artifacts by
extracting targets with a single greedy match (.*rm ...), which only ever
inspects the LAST rm in the command. A chain like 'rm -rf /; rm -rf
node_modules' was therefore judged solely by its trailing safe target and
allowed without warning, waving through the destructive 'rm -rf /'.

Gate the shortcut to single rm invocations: when any shell separator
(; | & newline, incl. JSON-escaped \n/\r from the grep extraction path)
is present, fall through to the destructive-pattern check, which warns on
any recursive rm. Single-command artifact cleanups still allow.

Adds 3 regression tests covering semicolon and && chains in both orders.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* harden(careful): substitution separators + capital -R recursive flag (#2039)

Two residual fail-opens in the same guard PR #2040 hardened, both verified
by executing the script pre-fix:

- rm -rf $(./wipe-all)/node_modules silently allowed: the substitution token
  ends in a whitelisted suffix and the safe-exception early exit skipped ALL
  downstream checks. $( and backtick now count as chain separators; plain
  $VAR expansion stays allowed.
- rm -R / silently allowed: both greps required a lowercase r in the flag
  cluster; capital -R is the documented BSD/macOS recursive flag. Both greps
  now match -[a-zA-Z]*[rR].

Six new tests: substitution x2 -> ask, capital-R x2 -> ask, rm -Rf
node_modules single-command -> still allowed, escaped-newline branch
(existing code, previously untested), and a pinned deliberate FP
(cd app && rm -rf node_modules -> ask) documenting the fail-closed
direction on chains.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(context-restore): prefer the current branch's own checkpoint (#2052)

All worktrees of a repo share one origin-derived slug, so they share one
`~/.gstack/projects/<slug>/checkpoints/` dir. `/context-restore` loaded the
newest checkpoint across the whole dir, so in one worktree it could silently
restore a *sibling worktree's* newer checkpoint.

Step 1 now orders candidates current-branch-first (read from each file's
`branch:` frontmatter), keeping other branches as a fallback. A branch is
checked out in at most one worktree, so this stops cross-worktree contamination
while preserving Conductor cross-branch handoff: when the current branch has no
checkpoint of its own, the full newest-first set is still used.

- scan the 200 newest before partitioning so a current-branch checkpoint sitting
  below a burst of sibling saves is still found; output still capped at 20
- non-git / detached HEAD / branchless legacy saves fall back to the old
  newest-first behavior (back-compat)
- +5 regression tests in context-save-hardening.test.ts (the #2052 bug case
  fails on the old pipeline); regenerated SKILL.md + proactive-suggestions.json

Fixes #2052

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(gbrain): pass --confirm-destructive on drift re-register (#1985)

ensureSourceRegistered() handles match-but-different-path by removing the
old source then re-adding it at the new path. The remove was issued as
`gbrain sources remove <id> --yes`, but gbrain >= 0.42 gates `sources
remove` behind `--confirm-destructive` (`--yes` alone no longer suppresses
the data-loss prompt). The remove therefore fails with "To proceed, pass
--confirm-destructive", which ensureSourceRegistered surfaces as "source
registration failed" — aborting the entire /sync-gbrain code stage for any
already-registered source whose path has drifted. The memory and brain-sync
stages still pass, so the code index silently stops refreshing.

The orchestrator's own safeSourcesRemove() already passes
--confirm-destructive; this brings the lib helper in line with that
convention. Keeps --yes for older gbrain.

Tests: extend the fake gbrain shim in gbrain-sources.test.ts to simulate
the gbrain >= 0.42 guard (remove without --confirm-destructive exits 1),
update the drift re-register assertion, and add a regression test that
proves the drift path no longer throws. Both fail on main with the exact
"To proceed, pass --confirm-destructive" error and pass with the fix.

Fixes #1985

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* harden(gbrain-sources): route drift remove through #1734 guards + realpath drift check

Absorbing #2031 un-blocked a destructive remove that bypassed the #1734
data-loss guards: ensureSourceRegistered's drift path issued
`gbrain sources remove` directly, without the detectAutopilot +
decideSourceRemove checks every other remove routes through via
safeSourcesRemove. gbrain >= 0.42's own prompt was accidentally blocking
that path; with --confirm-destructive passed it is live again.

- Drift remove now refuses LOUDLY (throws, actionable message) while an
  autopilot is active or when decideSourceRemove disallows; a silent
  changed=false would hide the drifted registration.
- decideSourceRemove's extraArgs (--keep-storage when supported) propagate
  to the remove call, matching safeSourcesRemove.
- Drift is realpath-normalized before being declared: a symlink alias of the
  same directory (macOS /tmp -> /private/tmp) is a match, not drift — the
  probable cause of #1985's reporter hitting the remove on an unmoved repo.
- Drift fires a loud stderr line (old -> new path); perpetual drift in logs
  is the trigger for promoting #1985's reindex-in-place design.

Tests: autopilot-active refusal (no remove in call log), fail-closed refusal
on unreadable sources list, --keep-storage propagation, symlink-alias
no-drift; existing drift tests pin the guard probes so a live autopilot on
the dev machine can't flip them.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(developer-profile): exclude mode:resources rows from SESSION_COUNT, TIER, NUDGE_ELIGIBLE (#2067)

Every /office-hours run appends a mode:"resources" bookkeeping row alongside
the real session row, so --read double-counted sessions (~2x): tiers promoted
early and the builder-to-founder nudge armed prematurely. The file already
filtered resources rows for LAST_*/CROSS_PROJECT; the same realSessions
filter now feeds SESSION_COUNT/TIER, and the nudge predicate is the faithful
allowlist (mode === 'builder') so a future mode #4 fails closed instead of
re-opening this bug.

8 regression tests: count vs resources noise, tier boundaries both sides,
nudge false-with-noise / true-at-3-builders, cross-project trailing row.

Absorbed from PR #1991 by @mvann (fix + tests commits; the PR's version-bump
commit is superseded by this wave's consolidated release commit).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(hooks): passThrough() two-branch contract — never emit permissionDecision:'defer' (#2035, #2006)

Every AskUserQuestion died with "Tool result missing due to internal error"
on current Claude Code builds (Desktop 1.14271.0, CC 2.1.177). Root cause:
the question-preference-hook emitted permissionDecision:'defer' on every
pass-through path. 'defer' is a real PreToolUse value, but since CC v2.1.89
its semantics are "pause this tool call for external resumption" (headless
resume) — never "abstain". Interactive sessions have nothing to resume the
paused call, so the tool orphaned. Pre-2.1.89 builds ignored the unknown
value, which is why the hook worked when it shipped and broke later.

The fix is the two-branch pass-through contract:
- no context -> exit 0 with EXACTLY empty stdout
- memory nuggets present -> hookSpecificOutput with hookEventName +
  additionalContext ONLY (the documented shape; plan-tune Layer 8 memory
  injection ships through this branch and keeps working)

defer() is renamed passThrough() so the function says what it does, and
docs/spikes/claude-code-hook-mutation.md's protocol contract (cited by the
hook header) is corrected in the same commit — it taught '"defer" — let
permission flow continue' and was the reintroduction vector.

Test contract rewritten in the same commit (13 assertions across 3 files,
verified fail-first against the unfixed hook): pass-through paths assert
exact-empty stdout (a garbage/partial write cannot slip past an
optional-chained parse), the nugget path asserts permissionDecision is
ABSENT while additionalContext survives, and a new tripwire asserts no
non-deny path ever puts the string "permissionDecision" on stdout. The
deny (auto-decide) and Conductor prose-redirect paths are unchanged.

Deployment: no migration needed — settings.json points at the absolute
bash shim which execs the .ts live; /gstack-upgrade delivers the fix.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(one-way-doors): unify credential noun net + wire it into the runtime (#2024)

Library fix: revoke/reset/rotate now share ONE noun alternation (api key,
token, secret, credential, access key, password) with optional plural s?.
Pre-fix leaks: "reset my secret", "reset my access key", "revoke my secret"
(mismatched per-verb lists) and every plural form ("rotate the credentials",
"revoke all tokens" — \b(...)\b cannot match a trailing s).

Runtime wiring — the regexes could never fire in production before:
- gstack-question-preference --check gains --summary-stdin: the question
  text pipes via stdin (never argv — summaries carry quotes/newlines/shell
  metacharacters) and feeds isOneWayDoor alongside the id, so an ad-hoc
  destructive question with a stored never-ask preference now forces
  ASK_NORMALLY. Empty/absent stdin keeps exact id-only semantics.
- question-preference-hook falls back to classifyQuestion(question text)
  when the registry lookup misses, so unregistered destructive questions
  pass through to a human instead of auto-deciding.
- question-tuning resolver prose shows the piped form (SKILL.md regen lands
  in the wave's release commit).

Tripwires (verified fail-first): full verbs x nouns x singular/plural matrix
with the #2024 repro rows, benign-summary no-over-match rows, stdin
transport survival (quotes/newlines), empty-stdin fail-safe, and hook
fallback both directions (destructive -> pass-through, benign -> deny).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(design): loud integer-flag contract for --count/--retry/--timeout (#2032)

design variants --count abc silently generated ZERO variants and exited 0:
parseInt(NaN) flowed through Math.min into the generation loop bound. The
same NaN class was live on the two sibling flags in the same file:
--retry abc made generate() a silent no-op (attempt <= NaN never true, null
output, exit 0) and --timeout abc killed the serve board ~immediately
(setTimeout(NaN)).

New design/src/flag-utils.ts: parseIntFlag (pure, unit-testable) +
normalizeIntFlag (CLI wrapper). Contract matches the --viewports precedent
(error loudly on nonsense — these commands spend real image-API money, a
silent fixup hides typos from calling agents): undefined -> default; bare
flag/empty/non-integer ("3.7" rejected, not truncated)/below-min -> exit 1
with usage hint; above-max -> clamp with stderr warning. --count normalizes
at the variants() consumption site so programmatic callers are covered, with
the ceiling derived from STYLE_VARIATIONS.length instead of a magic 7; the
CLI passes the raw flag through (a pre-parseInt would truncate "3.7").

Tripwires live in test/design-flag-utils.test.ts — deliberately under test/,
not design/test/, which is invisible to the bun test glob, TEST_ROOTS, and
every workflow (wiring design/test/ into CI is a captured TODO).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(gbrain): thin-client state — remote-MCP brains no longer classify as broken-config (#2051)

A thin client (remote-HTTP MCP brain, no local engine by design) probed
`gbrain sources list`, which gbrain's dispatch guard REFUSES on thin clients
(exit 1, no recognized error string), so the classifier fell to its
defensive broken-config default and every suppression gate silently hid
brain-aware blocks from exactly the users on a shared team brain.

New 'thin-client' state, detected PRE-probe from gbrain's own remote_mcp
config marker via the existing gbrainConfigPath() helper (mirrors gbrain's
isThinClient(); honors GBRAIN_HOME; zero network, immune to error-string
drift), with a /thin[- ]client/ stderr backstop in the probe catch. Remote
reachability is deliberately NOT probed by the classifier — that is the
#1964 pathology; gbrain calls degrade gracefully at use time, and the detect
JSON says so honestly (gbrain_thin_client: {probed: false}).

The state is admitted at every suppression gate — gstack-gbrain-detect
--is-ok (drives setup + gbrain-refresh), gen-skill-docs' detection override,
gstack-config gbrain-refresh — while the sync stages (code/memory/dream)
SKIP with an accurate reason: code indexing runs on the brain server, memory
syncs via the remote brain's artifacts pull. The two consumer classes need
opposite answers, which is why this is a distinct state and not a
skip-the-probe special case. sync-gbrain Step 1.5 and setup-gbrain prose
route thin-client to proceed, never into broken-config remediation.

detectMcpMode secondary generalization: url-match against the config's
remote_mcp.mcp_url (deterministic — gbrain mounts at the generic /mcp path)
-> name pattern gbrain[-_]* -> stdio command token; gbrain_mcp_mode stays a
3-value enum.

Tripwires: end-to-end --is-ok exits 0 on a thin-client fixture AND still
exits 1 on broken-config (the gate didn't widen); pre-probe + stderr-fallback
classifier paths; 4 detectMcpMode identification cases incl. a non-matching
url that must NOT false-positive.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* release: v1.60.0.0 — regen SKILL.md, VERSION, CHANGELOG, TODOS follow-ups

- Regenerate all SKILL.md from templates (question-tuning --summary-stdin
  prose from #2024, context-restore branch preference from PR #2054,
  sync-gbrain/setup-gbrain thin-client prose from #2051) + llms.txt.
- VERSION + package.json -> 1.60.0.0 (bin/gstack-next-version, queue-aware:
  #1815 claims 1.59.0.0, #2213 claims 1.59.1.0).
- CHANGELOG release summary + itemized entry crediting @jbetala7 (x3) and
  @mvann.
- TODOS.md: three eng-review follow-ups (design/test CI wiring + documented
  pre-existing retry-after flake, /context-save worktree identity, gbrain
  reindex-in-place conditional on the new drift log).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(resolvers): compress --summary-stdin preamble prose to fit parity budget; re-bless ship goldens

The v1.57.7.0 parity suite caps investigate's generated size at 1.09x
baseline; the #2024 question-tuning prose (duplicated into every tier->=2
skill) tipped it to 1.092. Compressed to a single inline command + short
pointer (the full rationale lives in bin/gstack-question-preference's
header and the one-way-doors module docs). Ship goldens re-blessed against
the final resolver text (conscious template-change acknowledgment, per the
golden-file regression contract).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(e2e): office-hours-spec-review turn budget fits the carved skill layout (#2473)

The test failed deterministically with error_max_turns at 9 turns on main
and this branch alike (CI attempt logs + local main repro). Root cause from
the failing transcript: the Spec Review Loop content is carved out of
office-hours/SKILL.md into office-hours/sections/, so the agent needs
discovery hops (grep SKILL.md -> ls sections/ -> read the section) before it
can write — 8 tool turns + the closing text turn = 9 > the 8-turn budget,
which predates the carve. Observed failures wrote a CORRECT summary on tool
turn 8 and died on the closing turn.

maxTurns 8 -> 12. Verified: PASS locally post-fix (7 turns this run — the
extra headroom absorbs discovery-path nondeterminism).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(e2e): review-dashboard-via session budget survives runner contention (#2473)

The test failed on CI (and its baseline run) with the timeout signature:
0 turns, $0.00, exactly 183s, 3/3 attempts — the spawned claude -p session
never emitted a single stream event before the 180s inner timeout. The
file's tests run concurrently on one runner; session startup queues behind
sibling sessions, and this test had the tightest budget in the file (the
240s-budget tests in the same job passed). A clean local run takes 270s
wall for 4 turns, confirming 180s was too tight even without contention.

Inner timeout 180s -> 300s; outer bun timeout 240s -> 360s to keep headroom
over the inner budget. Verified: PASS locally post-fix (4 turns, 270s).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(e2e): retro-base-branch session budget survives runner contention (#2473)

Same class as review-dashboard-via, one test over in the same file: /retro
is a long multi-step flow whose clean pass measures 225-239s — a coin flip
against the 240s inner budget. First CI run passed at 225s; the rerun timed
out at the 240s line on all 3 attempts (exitReason "timeout"); the local
verification run passed at 239s, ONE second under the old cap.

Inner timeout 240s -> 360s; outer bun timeout 300s -> 480s for headroom.
Verified: PASS locally post-fix (17 turns, 239s).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: Jayesh Betala <jayesh.betala7@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Michael Vann <9221873+mvann@users.noreply.github.com>
2026-08-08 09:28:45 -07:00

512 lines
19 KiB
TypeScript

#!/usr/bin/env bun
/**
* PreToolUse hook for AskUserQuestion (Claude Code, plan-tune cathedral T6).
*
* Enforces never-ask / always-ask / ask-only-for-one-way preferences
* deterministically — no agent compliance required.
*
* Decision tree (per question in tool_input.questions):
* 1. Extract question_id via marker (<gstack-qid:foo-bar>). If no marker,
* enforcement is skipped for this question (D18 — hash IDs are
* observed-only, never used as preference keys).
* 2. Look up door_type from scripts/question-registry.ts (default two-way).
* 3. Read preferences with precedence: project-local > global (D8).
* 4. Apply:
* never-ask + one-way → pass through (safety override; one-way always asks).
* never-ask + two-way + marker → deny with auto-decided recommendation
* in reason. Mark tool_use_id so PostToolUse logs as 'auto-decided'.
* ask-only-for-one-way + two-way + marker → same as never-ask.
* always-ask, or no preference → pass through.
*
* Pass-through = exit 0 with empty stdout (or additionalContext-only output
* when memory nuggets exist) — NEVER permissionDecision:'defer', whose
* CC v2.1.89+ semantics are pause-for-external-resumption (#2035, #2006).
*
* Why deny+reason instead of allow+updatedInput:
* AskUserQuestion's `updatedInput` shape for "pre-resolve this question"
* isn't structurally pinned in Claude Code docs (spike T4 left as open
* question). `deny` with a reason that names the auto-decided option is
* conservative + reliable: the model receives the rejection feedback,
* reads the recommended option from the reason, and proceeds without
* re-firing AUQ. When the spike around input mutation lands, we can
* swap to allow+updatedInput without changing the contract.
*
* Recommended-option extraction (per D2):
* - First: (recommended) label suffix on an option.
* - Fall back: "Recommendation: X" prose match against option labels.
* - Refuse to auto-decide if ambiguous (multiple labels OR no parseable
* recommendation): pass through instead of silent-wrong.
*
* Always exits 0. Hook errors land in ~/.gstack/hook-errors.log.
* See docs/spikes/claude-code-hook-mutation.md for the protocol contract.
*/
import * as fs from 'fs';
import * as path from 'path';
import * as os from 'os';
import { spawnSync } from 'child_process';
import { isConductor } from '../../../lib/is-conductor';
import { classifyQuestion } from '../../../scripts/one-way-doors';
interface HookStdin {
session_id?: string;
hook_event_name?: string;
tool_name?: string;
tool_use_id?: string;
tool_input?: {
questions?: Array<{
question?: string;
options?: Array<string | { label?: string; description?: string }>;
multiSelect?: boolean;
}>;
};
cwd?: string;
}
const MARKER_RE = /<gstack-qid:([a-z0-9-]{1,64})>/i;
const RECOMMENDED_LABEL_RE = /\(recommended\)\s*$/i;
function stateRoot(): string {
return (
process.env.GSTACK_STATE_ROOT ||
process.env.GSTACK_HOME ||
path.join(os.homedir(), '.gstack')
);
}
function logHookError(msg: string): void {
try {
const sr = stateRoot();
fs.mkdirSync(sr, { recursive: true });
fs.appendFileSync(
path.join(sr, 'hook-errors.log'),
`${new Date().toISOString()} question-preference-hook: ${msg}\n`,
);
} catch {
// last-resort swallow
}
}
function readStdin(): Promise<string> {
return new Promise((resolve) => {
let buf = '';
process.stdin.setEncoding('utf-8');
process.stdin.on('data', (chunk) => (buf += chunk));
process.stdin.on('end', () => resolve(buf));
process.stdin.on('error', () => resolve(buf));
setTimeout(() => resolve(buf), 2000);
});
}
function passThrough(additionalContext?: string): void {
// Abstain = exit 0 with EMPTY stdout (#2035, #2006). Never emit a
// permissionDecision here: 'defer' is a real PreToolUse value, but since
// Claude Code v2.1.89 its semantics are "pause this tool call for external
// resumption" (a headless-resume feature) — NOT "no opinion". In an
// interactive session nothing resumes the paused call, so every
// AskUserQuestion died with "Tool result missing due to internal error".
// additionalContext-only hookSpecificOutput is the documented shape for
// injecting context (plan-tune memory nuggets) without a decision.
if (additionalContext) {
process.stdout.write(
JSON.stringify({
hookSpecificOutput: {
hookEventName: 'PreToolUse',
additionalContext,
},
}),
);
}
process.exit(0);
}
function deny(reason: string): void {
process.stdout.write(
JSON.stringify({
hookSpecificOutput: {
hookEventName: 'PreToolUse',
permissionDecision: 'deny',
permissionDecisionReason: reason,
},
}),
);
process.exit(0);
}
function readJsonSafe(filePath: string): Record<string, unknown> | null {
try {
return JSON.parse(fs.readFileSync(filePath, 'utf-8'));
} catch {
return null;
}
}
interface PreferenceLookup {
preference: string | undefined;
source: 'project' | 'global' | 'none';
}
function lookupPreference(slug: string, questionId: string): PreferenceLookup {
const sr = stateRoot();
const projectFile = path.join(sr, 'projects', slug, 'question-preferences.json');
const globalFile = path.join(sr, 'global-question-preferences.json');
const project = readJsonSafe(projectFile);
if (project && typeof project[questionId] === 'string') {
return { preference: project[questionId] as string, source: 'project' };
}
const global = readJsonSafe(globalFile);
if (global && typeof global[questionId] === 'string') {
return { preference: global[questionId] as string, source: 'global' };
}
return { preference: undefined, source: 'none' };
}
interface RegistryEntry {
id: string;
door_type?: 'one-way' | 'two-way';
signal_key?: string;
}
interface MemoryNugget {
nugget: string;
applies_to_signal_keys: string[];
applied_at?: string;
}
/**
* Read per-session cache first, fall back to canonical local file. Cache
* invalidates by being missing — gstack-distill-apply doesn't touch the
* cache because the canonical file is always the source-of-truth on read
* miss. Sub-1ms cache reads (D13 perf).
*/
function loadMemoryNuggets(sessionId: string | undefined): MemoryNugget[] {
const sr = stateRoot();
const canonical = path.join(sr, 'free-text-memory.json');
let nuggets: MemoryNugget[] | null = null;
if (sessionId) {
const cachePath = path.join(sr, 'sessions', sessionId, 'memory-cache.json');
try {
const cached = JSON.parse(fs.readFileSync(cachePath, 'utf-8'));
if (Array.isArray(cached.nuggets)) {
return cached.nuggets;
}
} catch {
// miss → fall through
}
}
try {
const j = JSON.parse(fs.readFileSync(canonical, 'utf-8'));
nuggets = Array.isArray(j.nuggets) ? j.nuggets : [];
} catch {
nuggets = [];
}
// Write through to the per-session cache so subsequent hooks on this
// session take the fast path. Best-effort; never fails the hook.
if (sessionId && nuggets) {
try {
const dir = path.join(sr, 'sessions', sessionId);
fs.mkdirSync(dir, { recursive: true });
fs.writeFileSync(
path.join(dir, 'memory-cache.json'),
JSON.stringify({ nuggets, cached_at: new Date().toISOString() }, null, 2),
);
} catch {
// swallow
}
}
return nuggets || [];
}
/**
* For a given signal_key, return up to N nuggets whose applies_to_signal_keys
* include it. Sorted by recency (most-recently-applied first), capped.
*/
function nuggetsForSignal(nuggets: MemoryNugget[], signalKey: string, max = 3): string[] {
return nuggets
.filter((n) => Array.isArray(n.applies_to_signal_keys) && n.applies_to_signal_keys.includes(signalKey))
.sort((a, b) => (b.applied_at || '').localeCompare(a.applied_at || ''))
.slice(0, max)
.map((n) => n.nugget);
}
let registryCache: Record<string, RegistryEntry> | null = null;
function loadRegistry(): Record<string, RegistryEntry> {
if (registryCache) return registryCache;
registryCache = {};
try {
// Hook lives at hosts/claude/hooks/; registry at scripts/question-registry.ts
const here = path.dirname(new URL(import.meta.url).pathname);
const repoRoot = path.resolve(here, '..', '..', '..');
const regPath = path.join(repoRoot, 'scripts', 'question-registry.ts');
if (!fs.existsSync(regPath)) return registryCache;
const src = fs.readFileSync(regPath, 'utf-8');
// Cheap regex extraction so the hook doesn't need to import the TS file
// (which would require bun resolving the module at hook-invocation time).
// Matches entries like:
// 'ship-test-failure-triage': {
// id: 'ship-test-failure-triage',
// ...
// door_type: 'one-way',
// signal_key: 'test-discipline',
// ...
// },
const blockRe =
/'([a-z0-9-]+)':\s*\{[^}]*?door_type:\s*'(one-way|two-way)'[^}]*?\}/g;
let m: RegExpExecArray | null;
while ((m = blockRe.exec(src))) {
const [block, id, door_type] = m;
const sk = block.match(/signal_key:\s*'([a-z0-9-]+)'/);
registryCache[id] = {
id,
door_type: door_type as 'one-way' | 'two-way',
signal_key: sk ? sk[1] : undefined,
};
}
} catch (e) {
logHookError(`registry load failed: ${(e as Error).message}`);
}
return registryCache;
}
function optionLabels(opts: Array<string | { label?: string; description?: string }>): string[] {
return opts.map((o) => (typeof o === 'string' ? o : o.label || o.description || ''));
}
function extractRecommended(
questionText: string,
opts: string[],
): { recommended: string | undefined; ambiguous: boolean } {
const labelMatches = opts.filter((o) => RECOMMENDED_LABEL_RE.test(o));
if (labelMatches.length === 1) {
return { recommended: labelMatches[0].replace(RECOMMENDED_LABEL_RE, '').trim(), ambiguous: false };
}
if (labelMatches.length > 1) return { recommended: undefined, ambiguous: true };
const m = questionText.match(/Recommendation:\s*([^\n]+)/i);
if (!m) return { recommended: undefined, ambiguous: false };
const recPhrase = m[1].trim();
const prefixMatches = opts.filter((o) =>
o.toLowerCase().startsWith(recPhrase.toLowerCase().slice(0, 12)),
);
if (prefixMatches.length === 1) return { recommended: prefixMatches[0], ambiguous: false };
if (prefixMatches.length > 1) return { recommended: undefined, ambiguous: true };
return { recommended: undefined, ambiguous: false };
}
function slugFromCwd(cwd: string | undefined): string {
// Mirror gstack-slug's basename fallback. The full slug resolver shells out
// to git, which is too expensive on a hot hook path; the basename is close
// enough for preference lookup (preferences are keyed by question_id, slug
// is just the directory bucket).
if (!cwd) return 'unknown';
return path.basename(cwd);
}
function markAutoDecided(sessionId: string | undefined, toolUseId: string | undefined): void {
if (!sessionId || !toolUseId) return;
try {
const sr = stateRoot();
const dir = path.join(sr, 'sessions', sessionId);
fs.mkdirSync(dir, { recursive: true });
fs.writeFileSync(path.join(dir, `.auto-decided-${toolUseId}`), '');
} catch (e) {
logHookError(`markAutoDecided failed: ${(e as Error).message}`);
}
}
/**
* Log an auto-decided event directly from PreToolUse, since `deny` prevents
* the tool from running and PostToolUse never fires. Without this, /plan-tune
* Recent auto-decisions would be blind to enforcement hits.
*/
function logAutoDecided(
questionId: string,
questionSummary: string,
recommended: string,
optionsCount: number,
sessionId: string | undefined,
toolUseId: string | undefined,
cwd: string | undefined,
): void {
try {
const here = path.dirname(new URL(import.meta.url).pathname);
const repoRoot = path.resolve(here, '..', '..', '..');
const bin = path.join(repoRoot, 'bin', 'gstack-question-log');
const payload: Record<string, unknown> = {
skill: 'unknown',
question_id: questionId,
question_summary: questionSummary.slice(0, 200),
options_count: optionsCount,
user_choice: recommended.slice(0, 64),
recommended: recommended.slice(0, 64),
source: 'auto-decided',
session_id: sessionId?.slice(0, 64),
tool_use_id: toolUseId?.slice(0, 128),
};
spawnSync(bin, [JSON.stringify(payload)], {
encoding: 'utf-8',
stdio: ['ignore', 'pipe', 'pipe'],
timeout: 3000,
// cwd of the originating tool call so gstack-slug resolves to the
// project the user is actually in, not the hook script's location.
cwd: cwd && fs.existsSync(cwd) ? cwd : undefined,
});
} catch (e) {
logHookError(`logAutoDecided failed: ${(e as Error).message}`);
}
}
async function main(): Promise<void> {
const raw = await readStdin();
if (!raw.trim()) {
passThrough();
return;
}
let stdin: HookStdin;
try {
stdin = JSON.parse(raw);
} catch (e) {
logHookError(`stdin parse failed: ${(e as Error).message}`);
passThrough();
return;
}
const toolName = stdin.tool_name || '';
if (
toolName !== 'AskUserQuestion' &&
!toolName.match(/^mcp__.+__AskUserQuestion$/)
) {
passThrough();
return;
}
const questions = stdin.tool_input?.questions || [];
if (questions.length === 0) {
passThrough();
return;
}
// For multi-question AUQ, enforcement is all-or-nothing per call:
// we deny only if ALL questions have marker + never-ask + safe door type.
// Mixed cases pass through so the user still gets to answer.
const registry = loadRegistry();
const slug = slugFromCwd(stdin.cwd);
const memoryNuggets = loadMemoryNuggets(stdin.session_id);
// Compute Layer 8 memory context inline: any nuggets matching the
// signal_keys of the questions in this AUQ get surfaced as additionalContext.
// This applies whether we pass through OR deny — gives the agent + user the
// relevant prior context either way.
const contextNuggets: string[] = [];
for (const q of questions) {
const qText = q.question || '';
const marker = qText.match(MARKER_RE);
if (!marker) continue;
const entry = registry[marker[1]];
if (!entry?.signal_key) continue;
const hits = nuggetsForSignal(memoryNuggets, entry.signal_key);
for (const h of hits) {
if (!contextNuggets.includes(h)) contextNuggets.push(h);
}
}
const memoryContext = contextNuggets.length
? '[plan-tune memory] Past answers suggest: ' + contextNuggets.join(' | ')
: undefined;
// Determine whether EVERY question is eligible for never-ask auto-decide.
// We deliberately do NOT early-return pass-through on the first ineligible question:
// a Conductor session still needs the [conductor] prose deny as a fallback,
// so we compute eligibility, then branch. memoryContext is preserved on every
// non-enforcing exit. (All-or-nothing per-call semantics are unchanged: any
// ineligible question makes the whole call not auto-decidable.)
const autoDecisions: Array<{ id: string; recommended: string }> = [];
let fullyAutoDecidable = true;
for (const q of questions) {
const qText = q.question || '';
const marker = qText.match(MARKER_RE);
if (!marker) { fullyAutoDecidable = false; break; }
const questionId = marker[1];
const pref = lookupPreference(slug, questionId);
if (!pref.preference || pref.preference === 'always-ask') { fullyAutoDecidable = false; break; }
const entry = registry[questionId];
let doorType: string = entry?.door_type || 'two-way';
if (!entry) {
// #2024: an unregistered id used to default straight to two-way without
// consulting the keyword net, so an ad-hoc DESTRUCTIVE question with a
// stored never-ask preference auto-decided. classifyQuestion is a pure
// regex pass over the question text; on any failure keep the default
// (enforcement still requires an explicit stored preference).
try {
if (classifyQuestion({ summary: qText.replace(MARKER_RE, '').trim() }).oneWay) {
doorType = 'one-way';
}
} catch (e) {
logHookError(`one-way classifier failed: ${(e as Error).message}`);
}
}
// Safety override — even never-ask doesn't bypass one-way doors.
if (doorType === 'one-way') { fullyAutoDecidable = false; break; }
const opts = optionLabels(q.options || []);
const { recommended, ambiguous } = extractRecommended(qText, opts);
// Refuse-on-ambiguous per D2 — fail safe.
if (!recommended || ambiguous) { fullyAutoDecidable = false; break; }
autoDecisions.push({ id: questionId, recommended });
}
if (fullyAutoDecidable && autoDecisions.length > 0) {
// All questions were eligible for enforcement.
markAutoDecided(stdin.session_id, stdin.tool_use_id);
// Log each auto-decided question now, since deny prevents PostToolUse from
// firing. /plan-tune Recent auto-decisions reads source=auto-decided events.
for (let i = 0; i < autoDecisions.length; i++) {
const d = autoDecisions[i];
const q = questions[i];
const qText = (q.question || '').replace(MARKER_RE, '').trim();
const opts = optionLabels(q.options || []);
logAutoDecided(d.id, qText, d.recommended, opts.length, stdin.session_id, stdin.tool_use_id, stdin.cwd);
}
const reasonLines = autoDecisions.map(
(d) =>
`[plan-tune auto-decide] ${d.id}${d.recommended} (your never-ask preference). Proceed with that option without re-prompting. Change with /plan-tune.`,
);
deny(reasonLines.join('\n'));
return;
}
// Not fully auto-decidable. In Conductor, AskUserQuestion is unreliable
// (native is disabled, the mcp__conductor__AskUserQuestion variant is flaky),
// so deny the tool and redirect to a prose decision brief. This is TRANSPORT
// AVOIDANCE, not preference enforcement: it fires regardless of marker,
// preference, or door type — including one-way doors, which must reach the
// human via prose rather than the unreliable tool.
if (isConductor()) {
const conductorReason =
'[conductor] AskUserQuestion is unreliable in Conductor (native disabled, MCP variant flaky). ' +
'Do NOT call AskUserQuestion (native or any mcp__*__AskUserQuestion). Render this decision as a ' +
'PROSE decision brief now: a D<N> label, an ELI10 of the issue, a Recommendation line, then one ' +
'paragraph per choice carrying its `(recommended)` marker and `Completeness: X/10`; tell the user ' +
'to reply with a letter, then STOP. For a one-way/destructive confirmation, require an explicit ' +
'typed confirmation and do NOT proceed on a vague reply. Capture the decision with gstack-question-log ' +
'(PostToolUse will not fire on a prose path).' +
(memoryContext ? `\n${memoryContext}` : '');
deny(conductorReason);
return;
}
passThrough(memoryContext);
}
main().catch((e) => {
logHookError(`main crash: ${(e as Error).message}`);
passThrough();
});