The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted external content from GitHub Pull Requests (titles, bodies, commit messages, and review comments), which could be used to influence the agent's behavior.
Ingestion points: Data is fetched via gh pr view in Phase 1 of the analysis process described in SKILL.md.
Boundary markers: The instructions do not define clear delimiters or specific instructions for the agent to ignore or isolate potentially malicious instructions embedded within the PR data.
Capability inventory: The skill has the ability to execute shell commands (gh) and write files (Markdown and SVG) to the local filesystem at docs/architecture/.
Sanitization: No explicit sanitization or escaping of external PR data is mentioned before it is used to generate diagrams or the final analysis document.
[COMMAND_EXECUTION]: The skill performs command-line operations using the gh (GitHub CLI) tool to fetch PR metadata, file lists, and diffs.
Evidence: SKILL.md specifies the use of gh pr view --json files,title,body,comments,reviews to gather information.
Input validation: The skill includes logic to validate PR inputs against GitHub URL patterns or #\d+ formats, providing a layer of protection against direct command injection via the PR identifier.

architecture-impact