The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of pull requests, diffs, and patches which could contain malicious instructions. Ingestion points: Defined in SKILL.md as PRs, diffs, merge requests, and patch stacks. Boundary markers: No explicit delimiters or instructions to ignore content within the ingested data are provided to the agent. Capability inventory: The skill is configured with access to tools including Bash, Write, Read, Grep, and Glob as listed in the frontmatter of SKILL.md. Sanitization: There are no instructions for sanitizing or validating the content of the external data before it is processed by the agent.

code-review