Commit Graph
34 Commits
Author SHA1 Message Date
haelyra 3fa6e30502 fix: preserve rollback version state and bounded runtime workflows 2026-10-11 11:45:41 -04:00
haelyra 43febb9ada fix: preserve opt-in GateGuard contract questions from PR 3497 2026-10-11 02:00:27 -04:00
montjeffrey f7f47679b6 fix(gateguard): make the external-contract question opt-in
- questionIdsFor and condensedHintFor take an optional
  { externalContract: true }; without it a code change that handles data
  asks data-schema only, and an unknown profile gets the original four
  questions and condensed hint.
- The hook reads GATEGUARD_EXTERNAL_CONTRACT_QUESTION with the same
  spellings as GATEGUARD_METRICS and passes it to the full denial, the
  condensed denial and the metrics record, so all three agree.
- The effectiveness harness's arm patch forwards the option, so an arm
  never drops it silently.

Docs: change-profile.md gains an External-contract question section and
qualifies "byte-identical to the fixed four questions"; design-notes.md,
SKILL.md (switch table, data rule, word-for-word set) and CHANGELOG say
the fifth question is opt-in; question-effectiveness.md and the harness
README note that the runner clears GATEGUARD_* variables, so a gate arm
from a current tree measures the gate without the question.
skill-triggers@1.json: registryDigest refreshed for the SKILL.md body
change; the description and triggers are unchanged.
2026-10-11 00:45:01 -04:00
montjeffrey 329ca4e11a docs(gateguard): document type filters, path patterns and lookups
- design-notes.md: new Type filters and Path patterns sections; Search
  scope extends the search-of-the-target rule to name lookups; Flag
  parsing lists the new value flags.
- change-profile.md: Unicode line terminators make an edit non-trivial.
- skills/gateguard/SKILL.md and CHANGELOG.md describe the same rules.
- skill-triggers@1.json: refresh registryDigest for the SKILL.md body
  change. The skill's description is unchanged, so triggers and
  triggersDigest are untouched, as in the plan-orchestrate refresh.
2026-10-10 23:38:48 -04:00
Jeffrey Montoya f19425dbf1 docs(gateguard): move change-profile notes into their own file 2026-09-30 00:23:08 -04:00
Jeffrey Montoya d4045c7370 docs(gateguard): document NotebookEdit, hard links and exempt precedence 2026-09-29 21:50:27 -04:00
Jeffrey Montoya 20eb2395e9 fix(gateguard): gate MultiEdit by its file_path and widen surface checks
Also refuses rg -z as read-only and never writes metrics through a symlink.
2026-09-29 21:50:26 -04:00
Jeffrey Montoya 6e677b008c fix(gateguard): judge comment-only edits against the file they change 2026-09-29 21:50:25 -04:00
Jeffrey Montoya 42e9072136 feat(gateguard): credit test files by their module stem
Drop the hardcoded baseline alias from the evaluation script and refresh the results.
2026-09-29 21:50:23 -04:00
Jeffrey Montoya 21ea5f96a8 feat(gateguard): profile shell scripts and Python module constants 2026-09-29 21:50:22 -04:00
Jeffrey Montoya 0301055327 feat(gateguard): build condensed code hints from the question ids 2026-09-29 21:50:21 -04:00
Jeffrey Montoya 4c035dd834 feat(gateguard): record opt-in decision metrics 2026-09-29 21:50:20 -04:00
Jeffrey Montoya 15d4f169d3 feat(gateguard): gate sensitive targets in subagents
A subagent's pass does not unlock the path for the parent.
2026-09-29 21:50:19 -04:00
Jeffrey Montoya 6f0fbb6bc9 feat(gateguard): keep session-keyed state for 8 idle hours
The project-directory fallback keeps the 30-minute window.
2026-09-29 21:50:18 -04:00
Jeffrey Montoya 77ddf9a792 feat(gateguard): pass read-only shell commands before the routine gate
The first command that can change something is still gated.
2026-09-29 21:50:17 -04:00
Jeffrey Montoya c3b2a7e627 feat(gateguard): name the closest search that did not count
One sanitized line in the denial with a reason code, from the cached turn scan.
2026-09-29 21:50:15 -04:00
Jeffrey Montoya ad0d9183ff feat(gateguard): pass comment and whitespace-only edits
The file stays unchecked, so the next code change is still gated; counted in trivial_allows.
2026-09-29 21:50:14 -04:00
Jeffrey Montoya 10f67715b1 feat(gateguard): ask code questions that fit the change
Stable question ids; local call sites instead of importers when no public
declaration is touched; data question only when the change handles data.
2026-09-29 21:50:13 -04:00
Jeffrey Montoya 442a6da150 test(gateguard): cover added sensitive key and credential names
List id_ecdsa, id_dsa, .netrc and .pgpass as sensitive in the skill docs.
2026-09-29 21:50:12 -04:00
Jeffrey Montoya 92b9d7c7be docs(gateguard): document first-touch classes, search credit, and denial cap 2026-09-29 21:50:11 -04:00
haelyra 0765f7a2ca feat(profiles): add opt-in Lean/Full and hybrid Auto selection 2026-09-27 16:46:07 -04:00
Affaan Mustafa db61d1c76a fix(gateguard): warn that parallel-batch siblings may already be applied (#3136)
A first-touch Edit/Write denial marks the file checked so the retry
passes. Sibling edits to the same file in the same parallel batch are
therefore judged against post-denial state and silently apply, leaving
the file in a state neither version intended.

Hooks see tool calls one at a time, so a batch-wide lock is not
possible. Instead make the partial application explicit: the Edit,
Write, MultiEdit, and condensed denials now name the file and warn
that other edits from the same batch may already have been applied,
and SKILL.md tells agents to send dependent edits sequentially and
re-read the file after a gated batch.
2026-09-18 18:37:01 -04:00
haelyra e0252df02f fix: scope GateGuard exemptions to the project
Address #2921 and complete the segment-anchoring direction in #2979. Preserve explicit absolute exemptions while denying accidental matches in unrelated projects.
2026-09-07 16:27:01 -04:00
Souptik Chakraborty c4253805b5 docs(gateguard): address review - split full-disable, pin glob semantics
CodeRabbit review on #2611, all four findings:

- GATEGUARD_DISABLED sat in a table introduced as 'these do not disable the gate'. Moved to its own full-disable section with ECC_GATEGUARD, and corrected the accepted values against ECC_DISABLE_VALUES (0/false/off/disabled/disable - the earlier draft would have implied 'no' works, which it does not).

- Documented that a leading **/ compiles to .*/ and so needs a preceding separator: verified by reproducing the hook's glob->regex translation, **/tests/** matches /repo/tests/foo.js but not a bare relative tests/foo.js. Docs now say so and the example carries both forms. Matcher behaviour deliberately unchanged - widening it is a behaviour change, not a docs fix.

- Reverse-drift check now compares documented names against the parsed env reads instead of hookSource.includes(), so a name surviving only in a comment or error string no longer satisfies it.

- readGateguardEnvNames builds one Set from collected matches instead of mutating via Set#add, per the repo's no-in-place-mutation guideline.
2026-08-29 14:55:13 -04:00
Souptik Chakraborty 4377ea1753 docs(gateguard): document the graduated gate controls
GateGuard reads five GATEGUARD_* environment variables that were absent
from skills/gateguard/SKILL.md, so the only discoverable escape hatch was
ECC_GATEGUARD=off - disabling the load-bearing destructive-Bash gate
along with the noisy ones (#2573).

Documented, with defaults and exact accepted values read from the hook:

- GATEGUARD_BASH_ROUTINE_DISABLED (was undocumented everywhere)
- GATEGUARD_EXEMPT_GLOBS (previously only in a 2.1.0 release note)
- GATEGUARD_BASH_EXTRA_DESTRUCTIVE (was undocumented)
- GATEGUARD_DISABLED (was undocumented)
- GATEGUARD_STATE_DIR (was undocumented; named in a runtime warning)
- GATEGUARD_FACT_FORCE_FULL_DENIALS (already documented; folded into the
  same table for one lookup point)

Adds tests/ci/gateguard-env-documented.test.js, which asserts every
GATEGUARD_* variable the hook reads appears in the skill doc, and that the
doc names no variable the hook has stopped reading. That surface test is
what found the three knobs beyond the two the issue reported.

Docs and test only; no hook behaviour changes.

Refs #2573
2026-08-29 14:55:13 -04:00
Jun ec4925135c fix(gateguard): finish tool-agnostic checklist across edit gate and SKILL.md copies (#2274)
b3268fef (#2272) made the write-gate "confirm no existing file" item
tool-agnostic in the JS hook, but the rest of the checklist surface still
names Glob/Grep. On hosts without those tools the agent still hits a dead
tool call on:
- the edit-gate "list importers" item in the hook (scripts/hooks/gateguard-fact-force.js)
- both checklist items in all three SKILL.md copies (en, ja-JP, zh-CN)

Apply the same wording b3268fef introduced — "(search the tree — Glob/Grep,
or find/grep via Bash)" — to those five remaining spots so the whole gate is
consistent. Prose-only; no logic change.

Follow-up to #2272 / b3268fef.
2026-06-29 18:38:24 -07:00
BERORINPO db7f2a6fd5 fix(skills): move top-level origin frontmatter key under metadata
The official Agent Skills spec (agentskills.io/specification) whitelists exactly
6 top-level frontmatter keys (name/description/license/compatibility/metadata/
allowed-tools). A top-level `origin` key fails the official validator
(anthropics/skills quick_validate.py ALLOWED_PROPERTIES; skills-ref validate).

This moves `origin: X` -> `metadata.origin: X` across the canonical skills/
tree, preserving each value verbatim. Frontmatter-only, minimal diff.

- 251 SKILL.md updated (242 new metadata block, 9 appended to existing metadata)
- origin values preserved verbatim (verified 251/251)
- YAML validated on all changed files
- scoped to canonical skills/ only (docs/<lang> translations + tool mirrors
  .cursor/.kiro/.agents left untouched; presumably regenerated from canonical)

Addresses #2233
2026-06-11 21:12:21 +09:00
Affaan Mustafa 6319c7d309 fix: stability batch — hook stdin truncation, Codex exa TOML, Stop hook JSON, GateGuard repetition (#2227)
* fix(hooks): fail open on oversized stdin instead of echoing truncated JSON (#2222)

run-with-flags.js capped stdin at 1MB but every fallthrough path still
echoed the truncated string to stdout. The harness parses hook stdout as
JSON, got a document cut mid-stream, and blocked the tool call — so any
Edit/Write with a >1MB hook payload was permanently blocked by every
registered pre-write hook, before ECC_HOOK_PROFILE / ECC_DISABLED_HOOKS
gating could run.

- Exit 0 with empty stdout (no opinion) when the stdin cap trips, before
  any echo or gating logic.
- Flush stdout via write callback before process.exit: exiting right
  after stdout.write() dropped everything past the ~64KB pipe buffer,
  cutting even sub-cap pass-through payloads mid-JSON.

Regression tests cover the enabled, disabled, and missing-arg paths for
oversized payloads plus full echo of sub-cap >64KB payloads.

* fix(codex): stop emitting invalid exa url entry, align merge with connector policy (#2224)

The Codex MCP merge declared exa with a url key, but Codex's
[mcp_servers.*] TOML schema is stdio-only — the url key makes the
entire config.toml fail to load, bricking both the codex CLI and the
desktop app. Every install/update re-injected the line because the
urlEntry branch treated the broken entry as present.

- ECC_SERVERS now emits only the current default set per
  docs/MCP-CONNECTOR-POLICY.md: chrome-devtools (stdio, command/args).
  Retired servers (supabase, playwright, context7, exa, github, memory,
  sequential-thinking) are never re-emitted; existing user-managed
  entries are untouched.
- The merge now repairs the exact ECC-emitted broken form (url-only
  exa entry) on every run so re-running the installer fixes broken
  configs instead of preserving them. User stdio exa entries
  (command + mcp-remote) are left alone.
- check-codex-global-state.sh requires chrome-devtools instead of the
  retired set, and flags url-only exa entries with a repair hint.

Tests cover repair, re-run idempotence, stdio-entry preservation, and
no-retired-server emission in add, update, dry-run, and disabled modes.

* fix(hooks): never echo truncated stdin from Stop hooks (#2090)

Stop hooks follow the ECC pass-through convention (echo stdin on
stdout), but every echoing Stop hook capped stdin and echoed the capped
string. The Stop payload carries last_assistant_message, so a long
final assistant message produced a JSON document cut mid-stream on
stdout, which the harness reports as 'Stop hook error: JSON validation
failed' across the whole Stop chain.

Reproduced: a Stop payload with a >64KB last_assistant_message run
through run-with-flags + cost-tracker emitted exactly 65536 bytes of
invalid JSON (cost-tracker capped stdin at 64KB — far below realistic
Stop payloads).

- cost-tracker: raise the cap to 1MB (matching all other hooks) and
  suppress the pass-through echo when stdin was truncated.
- check-console-log, stop-format-typecheck, desktop-notify: suppress
  the echo when stdin was truncated; flush stdout before process.exit
  so sub-cap payloads are not cut at the ~64KB pipe buffer.
- All hooks keep exiting 0 (fail-open); diagnostics go to stderr.

New stop-hooks-stdout test asserts the contract for every registered
Stop hook: stdout is empty or valid JSON, exit code 0 — for realistic
100KB payloads and oversized >1MB payloads, via the production runner
and via direct invocation. Updated the old hooks.test.js case that
codified the truncated-echo behavior.

* fix(hooks): dampen GateGuard fact-force repetition in long sessions (#2142)

In long autonomous sessions the fact-force gate produced 10+
near-identical 'state facts -> blocked -> restate -> retry' blocks in
one context window, which measurably raises the odds of the model
collapsing into a degenerate single-token repetition loop.

- Track a per-session fact_force_denials counter in GateGuard state
  (merged max across concurrent writers, reset with the session, robust
  to malformed on-disk values).
- The first GATEGUARD_FACT_FORCE_FULL_DENIALS denials (default 3) keep
  the full four-fact block; later denials emit a condensed single-line
  message that carries the denial ordinal, so consecutive denials are
  structurally different and never textually identical.
- True retries of the same target remain allowed without re-prompting
  (unchanged). Destructive-Bash and routine-Bash gates are unchanged,
  as are the ECC_GATEGUARD=off / ECC_DISABLED_HOOKS escape hatches.

Eight new tests cover budget counting, condensed format, ordinal
advancement, retry pass-through, env tuning, malformed state, MultiEdit
dampening, and destructive-gate exemption.

* fix(hooks): keep security hooks able to block on oversized stdin (#2222)

Refine the truncation fail-open: instead of skipping the hook entirely,
the runner now suppresses only its own raw-echo when stdin was
truncated. The hook still executes and receives the truncated flag
(run() context / ECC_HOOK_INPUT_TRUNCATED), so config-protection keeps
blocking truncated protected-config payloads (its test requires exit 2)
while pass-through hooks fail open with empty stdout as before.

* style: apply repo formatter to touched hook files
2026-06-11 00:31:33 -04:00
Affaan Mustafa cfe770a735 fix: add gateguard recovery escape hatch 2026-04-30 11:26:15 -04:00
Junming 20041294d9 fix(gateguard): rewrite routineBashMsg to use fact-presentation pattern (#1531)
* fix(gateguard): rewrite routineBashMsg to use fact-presentation pattern

The imperative 'Quote user's instruction verbatim. Then retry.' phrasing
triggers Claude Code's runtime anti-prompt-injection filter, deadlocking
the first Bash call of every session. The sibling gates (edit, write,
destructive) use multi-point fact-list framing that the runtime accepts.

Align routineBashMsg with that pattern to restore the gate's intended
behavior without changing run(), state schema, or any public API.

Closes #1530

* docs(gateguard): sync SKILL.md routine gate spec with new message format

CodeRabbit flagged that skills/gateguard/SKILL.md still described the
pre-fix imperative message. Update the Routine Bash Gate section to
match the numbered fact-list format used by the new routineBashMsg().
2026-04-21 18:02:16 -04:00
setoandClaude Opus 4.6 dd2962ee92 fix: 5 bugs + 2 tests from 3-agent deep bughunt
Bugs fixed:
- B1: JS gate messages still said "cat one real record" -> redacted/synthetic
- B2: Destructive bash key used 200-char truncation (collision bypass) -> SHA256 hash
- B3: sanitizePath only stripped \n\r -> now strips null bytes, bidi overrides, all control chars
- B4: Tool name matching was case-sensitive (latent bypass) -> lookup map normalization
- B5: SKILL.md Gate Types missing MultiEdit -> added with explanation

Tests added:
- T1: MultiEdit gate denies first unchecked file (CRITICAL - was untested)
- T2: MultiEdit allows after all files gated

11/11 tests pass.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-13 16:32:46 +09:00
setoandClaude Opus 4.6 8cd6378c81 fix: cubic-dev-ai round 3 — SKILL.md consistency
P2: Description now says "Edit/Write/Bash (including MultiEdit)"
    instead of listing MultiEdit as a separate top-level gate

P2: Write Gate and Anti-Patterns now use same "redacted or synthetic
    values" wording as Edit Gate (was still "cat one real record")

All 3 gate doc sections now consistent. 9/9 tests pass.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-13 16:19:01 +09:00
setoandClaude Opus 4.6 4dbed5ff5b fix: cubic-dev-ai round 2 — 3 issues across SKILL.md + pruning
P1: Gate message asked for raw production data records — changed to
    "redacted or synthetic values" to prevent sensitive data exfiltration

P2: SKILL.md description now includes MultiEdit (was missing after
    MultiEdit gate was added in previous commit)

P2: Session key pruning now caps __prefixed keys at 50 to prevent
    unbounded growth even in theoretical edge cases

9/9 tests pass.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-13 16:11:33 +09:00
setoandClaude Opus 4.6 5a03922934 feat(hooks,skills): add gateguard fact-forcing pre-action gate
A PreToolUse hook that forces Claude to investigate before editing.
Instead of self-evaluation ("are you sure?"), it demands concrete facts:
importers, public API, data schemas, user instruction.

A/B tested: +2.25 quality points (9.0 vs 6.75) across two independent tasks.

- scripts/hooks/gateguard-fact-force.js — standalone Node.js hook
- skills/gateguard/SKILL.md — skill documentation
- hooks/hooks.json — PreToolUse entries for Edit|Write and Bash

Full package with config: pip install gateguard-ai
Repo: https://github.com/zunoworks/gateguard

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-04-12 11:41:33 +09:00