The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes standard development commands including git fetch, git rebase, and git push --force-with-lease to maintain and synchronize the PR branch state.\n- [COMMAND_EXECUTION]: It invokes local project tooling via ./scripts/agents/tooling/agentTool.ts to perform GitHub API operations such as fetching PR metadata, managing review threads, and monitoring CI status.\n- [COMMAND_EXECUTION]: The skill runs pnpm install, pnpm test, and pnpm lint to verify code integrity and regenerate lock files during conflict resolution processes.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests and acts upon PR titles, bodies, and review comments which can be influenced by external actors.\n
Ingestion points: PR metadata and review threads are fetched via getPrInfo and getReviewThreads as described in SKILL.md.\n
Boundary markers: No explicit boundary markers or specific 'ignore instructions' warnings are identified when the agent processes the content of these threads.\n
Capability inventory: The agent has the capability to modify code, force-push to the repository, and update the state of GitHub security alerts via the agentTool.ts wrapper.\n
Sanitization: Although sanitizePrBody is utilized to extract issue numbers, there is no general sanitization or filtering logic applied to the review text before the agent processes it.

enter-merge-queue