The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external command-line tools for validation purposes, including gh pr diff for fetching pull request differences, and radon or gocyclo for code complexity audits.
[REMOTE_CODE_EXECUTION]: The skill implements a multi-judge consensus mechanism by spawning sub-agents through platform-native backends such as spawn_agent, TeamCreate (Claude), and OpenCode.
[DATA_EXPOSURE]: The skill reads from local agent-specific directories (.agents/planning-rules/, .agents/findings/, .agents/pre-mortem-checks/) to load context for the validation process.
[INDIRECT_PROMPT_INJECTION]: As a validator of external artifacts, the skill possesses an attack surface for indirect prompt injection, where instructions embedded within the analyzed files (e.g., markdown plans or PR diffs) could attempt to influence the judges' verdicts.
Ingestion points: Reads artifacts (paths/PR IDs), planning rules, and registry files.
Boundary markers: None explicitly defined in the execution logic; relies on structured prompts.
Capability inventory: File reading, sub-agent spawning, and execution of static analysis tools.
Sanitization: No explicit sanitization of input artifacts is mentioned prior to judge review.

validate