The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from existing GitHub Pull Requests (titles and bodies) via gh pr view. This creates an attack surface where a malicious PR description could attempt to influence the agent's behavior during the update/create workflow.\n
Ingestion points: The output of gh pr view --json url,title,body in SKILL.md (Step 6).\n
Boundary markers: The skill explicitly wraps untrusted content in <untrusted_pr_body> tags and provides a preamble instructing the agent to treat the content as data only and ignore any embedded instructions.\n
Capability inventory: The skill can execute powerful commands including git push, gh pr create, gh pr edit, and gh pr merge.\n
Sanitization: PR bodies are written to temporary files via mktemp and passed to the gh CLI using the --body-file flag, effectively preventing shell command injection via command-line arguments.\n- [COMMAND_EXECUTION]: The skill performs various shell operations using git and gh to manage branches and commits. It implements security best practices such as quoting shell variables and avoiding bulk staging (git add -A) to minimize the risk of accidental secret exposure or command injection.

ship-it