The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data (git diffs and logs) which could contain malicious instructions meant to influence the agent's behavior. * Ingestion points: Data enters through git diff and git log commands as described in the Procedure section of SKILL.md. * Boundary markers: XML-style tags () are used in Draft mode to help separate instructions from the data being analyzed. * Capability inventory: The skill possesses the capability to execute git add, git commit, and git push using the Bash tool. * Sanitization: The skill provides logic for the agent to pause and ask for confirmation if it detects secrets, half-finished work, or changes that appear unintentional.

commitkit