The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by fetching Pull Request comments from human teammates and various AI code-review bots (ingestion points: gh api calls in rules/comment-fetching.md). This data is used to generate file edits and drive automated Git and GitHub operations.
Boundary markers: The skill uses structured templates (templates/suggestion-pack.md) that wrap verbatim comment bodies in markdown blockquotes and fenced suggestion blocks to separate them from instructions.
Capability inventory: The system possesses significant capabilities including file system modification (Edit, Write), version control operations (git commit, git push), and GitHub API interactions to resolve threads (gh api / resolveReviewThread).
Sanitization: Robust validation is implemented in Phase 4 using a two-gate process: a /critical skill for adversarial pre-mortem analysis (where 'Must-fix' findings override implementation) and a /confidence skill to score the proposed changes. This tiered validation significantly reduces the risk of applying malicious or incorrect suggestions.
[COMMAND_EXECUTION]: The skill makes extensive use of CLI tools including git, gh (GitHub CLI), and gw (a worktree management utility) to automate its workflow. These tools are used within their intended functional scope to manage PR metadata, branches, and worktree isolation.
[EXTERNAL_DOWNLOADS]: The skill fetches data from GitHub's official API to retrieve pull request reviews and comments. It also integrates with a self-improvement loop using lorekit-memory to list and write procedural lessons across runs. These operations utilize well-known services and vendor-aligned resources (mthines/gw-tools).

implement-suggestion