The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external PR comments from AI bots to automatically generate and push code fixes. This represents an attack surface for indirect prompt injection where malicious instructions could be embedded in PR comments to influence the agent's behavior.
Ingestion points: The scripts/review-bot-snapshot.mjs script fetches inline, review, and issue comments from the GitHub API.
Boundary markers: Absent. The instructions do not specify delimiters or provide warnings to ignore embedded instructions within the ingested comments.
Capability inventory: The skill gives the agent the ability to write code, execute repository validation scripts (e.g., npm run test), and perform git push operations.
Sanitization: Absent. The skill does not perform validation or filtering on the retrieved comment bodies.
[COMMAND_EXECUTION]: The skill and its helper script execute various CLI tools for PR management and version control.
Execution points: Uses gh (GitHub CLI) for viewing PRs and fetching comments, git for commit/push operations, and node to run the bundled snapshot script.
Validation scripts: Dynamically resolves and executes repository scripts such as precommit, lint, and test using package managers like npm, yarn, pnpm, or bun.
Implementation: The scripts/review-bot-snapshot.mjs script uses node:child_process.spawnSync to execute commands, which is a standard implementation for this functionality.

address-review-bots