The Agent Skills Directory

COMMAND_EXECUTION (HIGH): The skill performs multiple destructive and external-facing shell operations, including git add, git commit, git push, and gh pr create. It specifically instructs the agent to bypass human-in-the-loop safety checks with directives such as "Do NOT ask user for confirmation
auto-generate and proceed" and "No confirmation prompts
'Ship' is the confirmation. Execute immediately."
INDIRECT_PROMPT_INJECTION (HIGH): The skill is vulnerable to instructions embedded in the data it processes.
Ingestion points: The skill reads untrusted repository content via git diff --stat, git diff --cached, and git log in SKILL.md.
Boundary markers: None. There are no delimiters or instructions to treat the ingested diff content as data rather than instructions.
Capability inventory: The skill has write/execute capabilities via git push and gh pr create as seen in Steps 4 and 5 of SKILL.md.
Sanitization: No sanitization is performed on the ingested diff content before it is processed by the LLM to generate commit messages and PR bodies.
PROMPT_INJECTION (MEDIUM): The skill contains 'Override' style instructions that attempt to redefine the agent's safety constraints, specifically targeting the elimination of confirmation prompts for significant actions.

ship