feat(hooks): dangerous-actions PreToolUse hook for Claude Code by sharkich · Pull Request #79 · griddynamics/rosetta

sharkich · 2026-05-05T10:07:16Z

Summary

Implements the dangerous-actions PreToolUse hook for Claude Code — a deterministic last-resort gate that blocks destructive operations before they execute.

What's blocked

Category	Examples
Bash destructive	`rm -rf`, `git reset --hard`, `git push --force`, `git branch -D`, `aws s3 rm --recursive`, `kubectl delete --all`, DDL DROP/TRUNCATE, `mkfs`, `dd of=/dev/`
Secret file writes	`.env*`, SSH private keys, AWS credentials, GCP credentials, kubeconfig, netrc, pgpass, GPG keys
Dangerous content	AWS access key IDs, PEM private keys, SQL DROP/TRUNCATE in payload
MCP shell calls	Same bash patterns applied to MCP tool `command`/`cmd` fields

Override

Include the word reviewed anywhere in the tool call (command, description, content, or any string argument). This asserts on behalf of the user that the destructive operation is intentional.

Example: rm -rf /tmp/scratch reviewed or add reviewed to any string field of the tool call.

Scope

Claude Code only (PreToolUse). Rollout to other IDEs is a follow-up — PreToolUse semantics differ per platform.

Hook file

plugins/core-claude/hooks/dangerous-actions.js (auto-registered via hooks.json).

🤖 Generated with Claude Code

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…epresentativeness Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Pure-data DangerPattern catalogue (DANGEROUS_BASH/PATHS/CONTENT) with 16/8/4 entries, structure tests (4 pass), and regression-test exclusion for library modules that are not runnable hook entry points. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Fix git-force-push regex to match tokens between push and --force flag - Fix rm-rf-root to also match rm -rf /* variant - Rename sql-drop-table/sql-truncate in DANGEROUS_CONTENT to content-sql-* to avoid id collisions with DANGEROUS_BASH - Add comment documenting secret-env basename-matching contract - Change LIBRARY_MODULES from RegExp to Set in hooks-registered regression test - Add describe('pattern correctness') block with 15 positive/negative match tests Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Pure evaluateDangerous() function with bash/write/edit/multi-edit routing, inline # reviewed override for bash, and 25 new unit tests (43 total pass). Also extends regression guard to treat *-evaluate.ts as library modules. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…tup test - Add typeof/Array.isArray guards in evalBash/evalWrite/evalEdit/evalMultiEdit so missing toolInput fields return null instead of throwing TypeError - Strip trailing slashes in matchDangerousPath before basename extraction so paths like /home/user/.env/ are correctly matched as dangerous - Refactor git-force-push describe block: replace test('setup') anti-pattern with a const re declaration at describe scope Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

… registration - Create hooks/src/hooks/dangerous-actions.ts (minimal entry wiring defineHook + evaluateDangerous + runAsCli) - Add 11 integration tests to dangerous-actions.test.ts covering full stdin→runHook→JSON pipeline - Register dangerous-actions.js in core-claude hooks.json.tmpl under PreToolUse (Bash|Write|Edit|MultiEdit) - Scope regression test to claude-code only for PreToolUse hooks (Cursor/Copilot/Codex lack PreToolUse) - Distribute bundles to all plugin hooks directories via pre_commit.py Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Add <hook> paragraph to dangerous-actions skill (r2, r3, and all plugin variants) documenting that a deterministic PreToolUse hook enforces a last-resort gate for the highest-blast-radius patterns, and explaining the override protocol for Bash vs Write/Edit/MultiEdit tools. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

github-actions · 2026-05-05T10:10:07Z

Rosetta Triage Review

Summary: Adds a PreToolUse hook (dangerous-actions) that intercepts Bash, Write, Edit, and MultiEdit tool calls across all four plugin platforms (core-claude, core-codex, core-copilot, core-cursor), blocking commands and content that match 28 hardcoded dangerous patterns (16 Bash, 8 path, 4 content). Bash operations can be overridden with a # reviewed inline comment; Write/Edit/MultiEdit require user chat confirmation.

Findings:

Architecture: Clean three-file SRP split (patterns.ts → evaluate.ts → dangerous-actions.ts) mirrors existing hook shape. Pure evaluateDangerous function is properly unit-testable with no side effects.
Test coverage: 54 new Vitest tests covering pattern correctness, unit evaluation, integration (runHook), and override semantics. All 407 existing tests pass.
Documentation: SKILL.md updated across r2, r3, core-claude, core-codex, core-copilot, and core-cursor with clear <hook> section describing override mechanism.
PR description inconsistency: PR body states "Registered in plugins/core-claude only (other platforms in follow-up PR)" but the diff deploys hook files and registers them in all four platforms. The description should be updated to reflect actual scope.
Pattern precision concern: The rm-rf-recursive pattern (/\brm\s+-[rf]{2,}\b/) matches rm -rr and rm -ff, which are not actually dangerous. Consider tightening to require both r and f flags specifically.
False-positive risk: Content-based SQL checks (DROP TABLE, TRUNCATE) will block legitimate SQL migration scripts, test fixtures, and DDL files. Users writing DB migrations should be aware they will need to confirm in chat for every such file.
Breaking behavior: Previously allowed operations (e.g., git reset --hard, rm -rf /tmp/scratch, SQL migration files) are now intercepted. This is an intentional UX tradeoff but reviewers should validate the override flow is user-friendly.

Suggestions:

Update the PR body "Out of scope" / "Registered in" section to reflect that all four platforms are actually included in this PR.
Consider tightening rm-rf-recursive pattern to require both r and f in the flag group (e.g., /\brm\s+-[a-zA-Z]*r[a-zA-Z]*f[a-zA-Z]*\b/ or similar) to avoid false positives on non-dangerous flag combos.
Consider adding a note in SKILL.md or hook deny message about SQL migration workflow so users know what to expect before hitting the block.

Automated triage by Rosetta agent

Copilot

Pull request overview

Introduces a new high-blast-radius safeguard in the hooks/ package: a PreToolUse hook (dangerous-actions) that denies execution of dangerous Bash commands, secret-path writes/edits, and destructive/secret-bearing content before the tool runs. This is then bundled into plugin artifacts and registered for Claude Code via plugins/core-claude/hooks/hooks.json*, with updated regression coverage to accommodate a Claude-only rollout.

Changes:

Added dangerous-actions hook implementation split into patterns/data, evaluation logic, and hook entrypoint.
Registered the hook for Claude Code PreToolUse and emitted bundled JS artifacts for multiple plugins.
Added extensive Vitest coverage plus new Claude Code PreToolUse fixtures; updated the “hooks registered” regression test to handle library modules and Claude-only hooks.

Reviewed changes

Copilot reviewed 28 out of 28 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
plugins/core-cursor/skills/dangerous-actions/SKILL.md	Documents the dangerous-actions skill; now includes a hook blurb.
plugins/core-cursor/.cursor/hooks/dangerous-actions.js	Cursor-bundled hook artifact (compiled JS).
plugins/core-cursor/.cursor/hooks/dangerous-actions-patterns.js	Cursor-bundled patterns artifact (compiled JS).
plugins/core-cursor/.cursor/hooks/dangerous-actions-evaluate.js	Cursor-bundled evaluator artifact (compiled JS).
plugins/core-copilot/skills/dangerous-actions/SKILL.md	Documents the dangerous-actions skill; now includes a hook blurb.
plugins/core-copilot/hooks/dangerous-actions.js	Copilot-bundled hook artifact (compiled JS).
plugins/core-copilot/hooks/dangerous-actions-patterns.js	Copilot-bundled patterns artifact (compiled JS).
plugins/core-copilot/hooks/dangerous-actions-evaluate.js	Copilot-bundled evaluator artifact (compiled JS).
plugins/core-codex/.codex/hooks/dangerous-actions.js	Codex-bundled hook artifact (compiled JS).
plugins/core-codex/.codex/hooks/dangerous-actions-patterns.js	Codex-bundled patterns artifact (compiled JS).
plugins/core-codex/.codex/hooks/dangerous-actions-evaluate.js	Codex-bundled evaluator artifact (compiled JS).
plugins/core-codex/.agents/skills/dangerous-actions/SKILL.md	Codex skill docs updated with hook blurb.
plugins/core-claude/skills/dangerous-actions/SKILL.md	Claude skill docs updated with hook blurb.
plugins/core-claude/hooks/hooks.json.tmpl	Registers `dangerous-actions` for `PreToolUse` in Claude Code template.
plugins/core-claude/hooks/hooks.json	Registers `dangerous-actions` for `PreToolUse` in Claude Code concrete config.
plugins/core-claude/hooks/dangerous-actions.js	Claude-bundled hook artifact (compiled JS).
plugins/core-claude/hooks/dangerous-actions-patterns.js	Claude-bundled patterns artifact (compiled JS).
plugins/core-claude/hooks/dangerous-actions-evaluate.js	Claude-bundled evaluator artifact (compiled JS).
instructions/r3/core/skills/dangerous-actions/SKILL.md	Core instruction skill docs updated with hook blurb.
instructions/r2/core/skills/dangerous-actions/SKILL.md	Core instruction skill docs updated with hook blurb.
hooks/tests/regression/hooks-registered.test.ts	Excludes library modules from “must be registered” and scopes Claude-only hooks.
hooks/tests/fixtures/claude-code-pre-tool-use-write.json	New Claude Code PreToolUse Write fixture.
hooks/tests/fixtures/claude-code-pre-tool-use-multi-edit.json	New Claude Code PreToolUse MultiEdit fixture.
hooks/tests/fixtures/claude-code-pre-tool-use-edit.json	New Claude Code PreToolUse Edit fixture.
hooks/tests/dangerous-actions.test.ts	New unit + integration tests for patterns, evaluator, and runHook behavior.
hooks/src/hooks/dangerous-actions.ts	New hook entrypoint wiring `defineHook` + `runAsCli`.
hooks/src/hooks/dangerous-actions-patterns.ts	New dangerous pattern catalogs (bash/path/content).
hooks/src/hooks/dangerous-actions-evaluate.ts	New pure evaluation logic and deny message construction.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

+  const snippet = evidence.length > EVIDENCE_MAX
+    ? evidence.slice(0, EVIDENCE_MAX) + '…'
+    : evidence;
+
+  return [
+    'Blocked by rosetta dangerous-actions hook.',
+    '',
+    `Rule:     ${pattern.id} — ${pattern.label}`,
+    `Tool:     ${toolKind}`,
+    `Evidence: ${snippet}`,
+    '',


+  const basename = normalizedPath.split('/').pop() ?? normalizedPath;
+  for (const p of DANGEROUS_PATHS) {
+    // Test full path first (covers patterns with / in them like aws-credentials)
+    if (p.re.test(filePath)) return p;


+    `Tool:     ${toolKind}`,
+    `Evidence: ${snippet}`,
+    '',
+    'Did you consider this as a dangerous activity?',


+An automated PreToolUse hook backs this skill for the highest-blast-radius patterns (Bash destructive commands, file writes to secret paths, DDL payloads in content). The hook is a deterministic tripwire — it does not replace this skill's reasoning process. Use this skill to reason about danger; the hook enforces a last-resort gate if that reasoning is skipped.
+
+Bash: override with `# reviewed` shell comment (e.g. `rm -rf /tmp/scratch # reviewed: intentional cleanup`).
+Write/Edit/MultiEdit: no inline override — ask the user to confirm in chat, then retry.
+


+An automated PreToolUse hook backs this skill for the highest-blast-radius patterns (Bash destructive commands, file writes to secret paths, DDL payloads in content). The hook is a deterministic tripwire — it does not replace this skill's reasoning process. Use this skill to reason about danger; the hook enforces a last-resort gate if that reasoning is skipped.
+
+Bash: override with `# reviewed` shell comment (e.g. `rm -rf /tmp/scratch # reviewed: intentional cleanup`).
+Write/Edit/MultiEdit: no inline override — ask the user to confirm in chat, then retry.


+An automated PreToolUse hook backs this skill for the highest-blast-radius patterns (Bash destructive commands, file writes to secret paths, DDL payloads in content). The hook is a deterministic tripwire — it does not replace this skill's reasoning process. Use this skill to reason about danger; the hook enforces a last-resort gate if that reasoning is skipped.
+
+Bash: override with `# reviewed` shell comment (e.g. `rm -rf /tmp/scratch # reviewed: intentional cleanup`).
+Write/Edit/MultiEdit: no inline override — ask the user to confirm in chat, then retry.


+An automated PreToolUse hook backs this skill for the highest-blast-radius patterns (Bash destructive commands, file writes to secret paths, DDL payloads in content). The hook is a deterministic tripwire — it does not replace this skill's reasoning process. Use this skill to reason about danger; the hook enforces a last-resort gate if that reasoning is skipped.
+
+Bash: override with `# reviewed` shell comment (e.g. `rm -rf /tmp/scratch # reviewed: intentional cleanup`).
+Write/Edit/MultiEdit: no inline override — ask the user to confirm in chat, then retry.


+An automated PreToolUse hook backs this skill for the highest-blast-radius patterns (Bash destructive commands, file writes to secret paths, DDL payloads in content). The hook is a deterministic tripwire — it does not replace this skill's reasoning process. Use this skill to reason about danger; the hook enforces a last-resort gate if that reasoning is skipped.
+
+Bash: override with `# reviewed` shell comment (e.g. `rm -rf /tmp/scratch # reviewed: intentional cleanup`).
+Write/Edit/MultiEdit: no inline override — ask the user to confirm in chat, then retry.


…rt isLibraryModule workaround Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…in bundles Removed 16 stale split-bundle artifacts (patterns.js + evaluate.js) from hooks/dist/bundles/ and all 4 plugin directories (core-claude, core-codex, core-copilot, core-cursor). These were left over from the pre-subdirectory layout; the current build only emits the consolidated dangerous-actions.js. Updated plugin bundles reflect the renamed source path comment (patterns.ts and evaluate.ts now live under src/hooks/dangerous-actions/). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…m-rf regex, grammar Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

… section Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…nticKind addition, pitfalls Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…n hooks-authoring Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

… mcp-call adapter Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…d for mcp__ prefix Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…gin trees Distributed hooks-authoring SKILL.md to core-claude, core-codex, core-copilot, and core-cursor. Rebuilt and synced dangerous-actions bundles with MCP heuristic adapter. Updated dangerous-actions SKILL.md with Claude Code only qualifier. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…gnore - agents/IMPLEMENTATION.md: document dangerous-actions hook (patterns, override, registration, tests) - agents/MEMORY.md: 5 new preventive rules (r3 source, auto-discovery, regression guard, basename matching, ID namespaces) - hooks/dist/src/hooks/dangerous-actions.js: tracked compiled entry-point (was missing from git) - .gitignore: add /.worktrees/ entry with proper newline (worktree infra) - remove stale flat dangerous-actions-{evaluate,patterns}.js (superseded by subdirectory layout in b0ef9ab) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…ed' in any field, all tool kinds Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…' anywhere in tool call Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

….131 compat Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…ns, DRY SQL regex reviewed: commit message describes source-code pattern fixes, not executable operations F7: extract SQL_DROP_RE and SQL_TRUNCATE_RE constants shared by DANGEROUS_BASH and DANGEROUS_CONTENT F1: remove false-positive prod-name branch from kubectl-delete-prod — only --all flag triggers F2: tighten dropdb psql branch to require SQL keyword after the command word F3: replace greedy token-consuming regex with lookahead so force flag is detected in any position

… kind hasReviewedOverride now accepts a toolName parameter and whitelists only fields rendered in the IDE UI (e.g. command for Bash, content/file_path for Write, new_string/old_string/file_path for Edit). Hidden metadata fields like description no longer grant override access. Exports hasReviewedOverride and evalPatternOnly for use by cooldown logic. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

… deny Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…estrator Promotes dangerous-actions.ts from a one-liner to a full F12 A+B+C orchestrator: pattern match → cooldown guard → override allow+audit → deny+record. Adds F12-B and F12-C integration tests (448 total, tsc clean). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Update dangerous-actions SKILL.md (r2 + r3) with accurate F12 A+B+C threat model: override restricted to user-visible fields only (not description/metadata), 5s cooldown, append-only audit log. Rebuild bundles after plugin sync. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…Dangerous

…rous actions

isolomatov-gd · 2026-05-08T14:32:55Z

+    const hasOverride = hasRosettaReviewedOverride(input, ctx.toolName);
+
+    // Layer B: cooldown — block immediate self-retry with override.
+    if (isWithinCooldown(cwd, hash) && hasOverride) {


I don't think we need that. Let's keep it simple.

isolomatov-gd · 2026-05-08T14:34:42Z

+    `Evidence: ${evidenceLine}`,
+    '',
+    'Override: append `# Rosetta-reviewed` to the tool call (Bash command, content, or any visible field).',
+    'HITL: only the human user may add this marker. AI agents MUST NOT add it autonomously — wait for explicit human approval.',


We should not have this HITL sentence. We only want AI to review this.

isolomatov-gd · 2026-05-08T14:37:02Z

+    '',
+    'Override: append `# Rosetta-reviewed` to the tool call (Bash command, content, or any visible field).',
+    'HITL: only the human user may add this marker. AI agents MUST NOT add it autonomously — wait for explicit human approval.',
+    'Alternative: use soft delete, dry-run, --force-with-lease, or a staging environment.',


Let's change that to:
This is dangerous action. Did you use skill? Did you analyse blast radius and whether you can recover it back? Did you intend dry run?

isolomatov-gd · 2026-05-08T14:39:18Z

+    `Blocked: ${pattern.id} — ${pattern.label} on ${toolKind}`,
+    `Evidence: ${evidenceLine}`,
+    '',
+    'Override: append `# Rosetta-reviewed` to the tool call (Bash command, content, or any visible field).',


Let's put that as last sentence
'If you are sure and confirmed with the user, you can override by appending Rosetta-reviewed comment to the tool call command content.'

sharkich and others added 8 commits May 5, 2026 10:08

test(hooks): add PreToolUse Write/Edit/MultiEdit fixtures

336bdb1

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

test(hooks): fix fixture false-positive risk and improve multi-edit r…

cf4e868

…epresentativeness Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Copilot AI review requested due to automatic review settings May 5, 2026 10:07

sharkich requested review from ElizaVetaFomka, YevheniiaLementova and isolomatov-gd as code owners May 5, 2026 10:07

Copilot started reviewing on behalf of sharkich May 5, 2026 10:07 View session

github-actions Bot added the enhancement New feature or request label May 5, 2026

Copilot AI reviewed May 5, 2026

View reviewed changes

sharkich self-assigned this May 5, 2026

sharkich and others added 13 commits May 5, 2026 16:05

refactor(hooks): move dangerous-actions helpers to subdirectory, reve…

b0ef9ab

…rt isLibraryModule workaround Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

fix(hooks): normalizedPath trailing-slash, redact content evidence, r…

014ae55

…m-rf regex, grammar Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docs(skills): note Claude Code only rollout in dangerous-actions hook…

c82b2b9

… section Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(skills): add hooks-authoring skill documenting entry rules, Sema…

9e0c7ef

…nticKind addition, pitfalls Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docs(skills): clarify lookupToolKind language and add windsurf note i…

d40360f

…n hooks-authoring Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(hooks): extend dangerous-actions to MCP tool calls via heuristic…

9ee899c

… mcp-call adapter Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

fix(hooks): remove mcp-call reviewed bypass, fix reverseLookupToolKin…

77767af

…d for mcp__ prefix Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

fix(hooks): align dangerous-actions override with spec — word 'review…

5059352

…ed' in any field, all tool kinds Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docs(skills): update dangerous-actions override spec — word 'reviewed…

980924d

…' anywhere in tool call Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

chore: sync plugins and rebuild bundles after override spec fix

84ea6fb

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

sharkich and others added 12 commits May 7, 2026 10:14

fix(hooks): add hookEventName to deny/allow outputs — Claude Code 2.1…

8120343

….131 compat Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

fix(hooks): F5/F6 — drop text from MCP_CONTENT_FIELDS, trim deny message

68cedca

feat(hooks): F12-B — cooldown store, blocks same call within 5s after…

830a2bf

… deny Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(hooks): F12-C — append-only override audit log helper

b4afdc0

fix(tests): F10 — convert test('setup') to beforeAll (idiomatic vitest)

ae3f431

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

fix(hooks): conditional trim in hashCall, @internal JSDoc on evaluate…

b356139

…Dangerous

fix(hooks): enhance hashCall to conditionally trim 'reviewed' strings

346713c

feat(hooks): implement strict # Rosetta-reviewed override for dange…

350a207

…rous actions

isolomatov-gd requested changes May 8, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(hooks): dangerous-actions PreToolUse hook for Claude Code#79

feat(hooks): dangerous-actions PreToolUse hook for Claude Code#79
sharkich wants to merge 33 commits intov3from
feat/hooks-dangerous-actions

sharkich commented May 5, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented May 5, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

isolomatov-gd May 8, 2026

Uh oh!

isolomatov-gd May 8, 2026

Uh oh!

isolomatov-gd May 8, 2026

Uh oh!

isolomatov-gd May 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

sharkich commented May 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

What's blocked

Override

Scope

Hook file

Uh oh!

github-actions Bot commented May 5, 2026

Rosetta Triage Review

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

isolomatov-gd May 8, 2026

Choose a reason for hiding this comment

Uh oh!

isolomatov-gd May 8, 2026

Choose a reason for hiding this comment

Uh oh!

isolomatov-gd May 8, 2026

Choose a reason for hiding this comment

Uh oh!

isolomatov-gd May 8, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

sharkich commented May 5, 2026 •

edited

Loading