The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data from GitHub issues and comments, creating a potential surface for indirect prompt injection.
Ingestion points: The skill reads live issue bodies and plan comments as defined in SKILL.md and references/executor-brief.md.
Boundary markers: The skill implements a robust boundary check via scripts/plan-header.mjs, which validates a requirementsSha256 fingerprint to ensure the plan matches the authorized issue requirements.
Capability inventory: The agent can perform Git operations, publish PRs, and execute local validation scripts.
Sanitization: The scripts/plan-header.mjs performs strict field validation, and the executor-brief.md mandates a scan for placeholders or TBDs before proceeding.
[COMMAND_EXECUTION]: The skill executes Git commands and local Node.js scripts for repository management and plan validation.
Evidence: SKILL.md and executor-brief.md utilize git fetch, git rev-parse, and node scripts/plan-header.mjs to maintain state consistency and verify authorization.
[SAFE]: The skill demonstrates security best practices, notably in references/delivery-gate.md, which requires that independent reviews come from a different model provider or lab than the implementation contributors to prevent systemic biases or bypasses.

executing-plans