The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes dynamic context injection placeholders (!command) in SKILL.md to execute git branch, git log, and git diff commands when the skill is loaded. These operations are restricted to gathering repository context for the pull request.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources that are then used to generate pull request titles and descriptions.
Ingestion points: The outputs of git log master..HEAD --oneline and git diff master...HEAD --stat are included in the agent context via dynamic injection in SKILL.md.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore potential injection patterns within the git history or diff data.
Capability inventory: The agent has the capability to execute git push and gh pr create based on the processed context.
Sanitization: There is no evidence of filtering or sanitization performed on the commit messages or code changes before they are processed by the agent.

pr