The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could be used to influence the agent's actions through embedded instructions.
Ingestion points: The skill reads issue bodies, comments, and pull request diffs from an external tracker (SKILL.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified when reading external content.
Capability inventory: The skill can perform file system operations (writing to .out-of-scope/, CONTEXT.md, and ADRs), execute shell commands (running tests), and interact with the GitHub API (posting comments, closing issues).
Sanitization: No sanitization or validation of the external content is mentioned before it is processed or used to update internal project files.
[DYNAMIC_EXECUTION]: The skill is instructed to execute code provided by external contributors to verify functionality.
Evidence: The SKILL.md file contains instructions to "check it out, run the relevant tests or commands" for pull requests to confirm their claims.
[COMMAND_EXECUTION]: The verification process involves running shell-based tests and commands, which are derived from the context of untrusted pull requests.

triage