The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data from the repository, which constitutes an injection surface. However, the risk is mitigated by the structured review process.
Ingestion points: The agent reads AGENTS.md, docs/, worktrees, and pull request data (SKILL.md, steps 1 and 2).
Boundary markers: No explicit prompt boundary markers are used in the instruction text.
Capability inventory: The skill utilizes apply_patch and references development tools such as gh (GitHub CLI) and cargo (Rust toolchain).
Sanitization: The INCIDENT_LOG.md indicates that the project maintains a 'hostile-data fallback matrix' and performs 'filesystem-security reviews' for processed context.
[COMMAND_EXECUTION]: The skill workflow involves executing standard development tools like gh and cargo to manage pull requests and verify code quality. These actions are aligned with the skill's primary purpose of software implementation.
[SAFE]: The inclusion of a detailed INCIDENT_LOG.md demonstrates a mature security posture. The log tracks and remediates process-level vulnerabilities, such as workflow-command injection and symlink traversal, ensuring the implementation process becomes more resilient over time.

implement