The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Node.js scripts (contract-snapshot.mjs, scan-diff-hygiene.mjs) and Git commands to analyze repository state and diffs. These operations are restricted to the local repository context and are necessary for the skill's stated purpose of code review and maintenance.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data, including pull request descriptions, comments, issues, and code diffs. It incorporates specific safeguards to mitigate injection risks:
Ingestion points: Untrusted data enters the agent context via gh pr view JSON output and file system reads of repository artifacts (e.g., in Phase 0 and 3).
Boundary markers: The instructions implement a multi-phase workflow (Phase 0-8) that explicitly separates "reviewed data" from "agent instructions." It specifies that repository content (comments, issues, diffs) should not change the agent's behavior.
Capability inventory: The skill possesses the capability to write files, execute Git commands, and run local Node.js scripts, which are leveraged during the "Deslop apply" operations in Phase 6.
Sanitization: The skill uses structured data parsing (JSON) for PR metadata and regex-based scanning via scan-diff-hygiene.mjs to evaluate repository content without direct execution.

review