The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external sources.
Ingestion points: Pull request review comments from humans and automated bots (e.g., Sentry, CodeRabbit) are retrieved via get_review_comments in references/assess-checklist.md.
Boundary markers: The instructions do not define explicit delimiters or warnings to ignore malicious instructions within the comments.
Capability inventory: The skill has the capability to write files (applying fixes), execute local scripts via npm, and perform VCS operations with git as described in references/fix-strategies.md.
Sanitization: While provider adapters normalize data, there is no evidence of sanitization or filtering of the raw comment bodies before they are used to inform code changes.
[COMMAND_EXECUTION]: The skill uses shell commands to perform standard development tasks within the local environment as detailed in references/fix-strategies.md.
VCS operations: Includes git checkout, git commit --amend, git rebase, and git push --force-with-lease.
Project scripts: Includes npm run lint, npm run test:run, npm run build, and npm run typecheck to validate code fixes.

shepherd