The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the GitHub (gh) and GitLab (glab) command-line interfaces to retrieve repository metadata, issue details, and pull request statuses. The implementation in scripts/fetch_issue_context.py uses subprocess.run with argument lists rather than shell strings, which prevents shell injection vulnerabilities. These operations are restricted to read-only queries by design.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external forge platforms (titles, descriptions, and comments). This creates a potential surface for indirect prompt injection where an attacker could place malicious instructions inside a GitHub issue. However, the skill implements strong structural mitigations:
Mutation Boundary: It strictly separates reading from writing; all destructive actions (like comments or PR creation) are returned as text payloads for human review instead of being executed by the agent.
Safe Interpolation: Instructions explicitly forbid the agent from using eval or interpolating untrusted text directly into shell commands, requiring the use of body files and proper argument quoting.
Input Validation: Repository names and issue numbers are parsed and validated using strict regular expressions before being passed to CLI tools.

rhdh-forge