The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted content from GitHub, including PR bodies, commit messages, and review comments, which may contain adversarial instructions intended to bias the agent's retrospective.
Ingestion points: Data is fetched in references/queries.md using commands like gh pr view for bodies, gh api for commit messages, and GraphQL queries for reviews and comments.
Boundary markers: The instructions do not define specific delimiters or instructions to isolate the external data from the agent's analytical instructions, which could lead to the agent following instructions found within the data.
Capability inventory: The skill's capabilities are limited to read-only GitHub API operations via the gh CLI; it does not possess the ability to write to the repository or execute scripts found in the content.
Sanitization: There is no evidence of sanitization or filtering applied to the retrieved GitHub content before it is processed by the model.

eg-github-contribution-retrospective