The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the $ARGUMENTS variable directly in shell commands and file paths, which could allow shell injection and path traversal.
Evidence: gh issue view $ARGUMENTS and gh issue comment $ARGUMENTS --body ... in SKILL.md interpolate unvalidated input into shell commands.
Evidence: docs/issues/issue-$ARGUMENTS.md in SKILL.md uses unvalidated input to construct a file write path.
Evidence: !git log --oneline -20 -- [relevant-paths] in SKILL.md utilizes dynamic context injection for repository analysis.
[PROMPT_INJECTION]: The skill processes untrusted data from GitHub issues, providing a surface for indirect prompt injection.
Ingestion points: Issue data (title, description, and comments) is fetched using the gh CLI in SKILL.md.
Boundary markers: There are no delimiters or instructions to treat the fetched content as untrusted data.
Capability inventory: The agent has access to shell commands and file system operations.
Sanitization: No sanitization or validation of retrieved content is performed.
[EXTERNAL_DOWNLOADS]: The skill interacts with GitHub's services to retrieve repository and issue data.
Evidence: Use of the GitHub CLI to download issue information from GitHub repositories.

piv-investigate-issue