The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to treat external, potentially untrusted files and logs as authoritative sources for its operational behavior. * Ingestion points: The skill reads instructions from AGENTS.md, contributing guides, and git log (SKILL.md). * Boundary markers: None; the skill explicitly states that project-specific conventions override its own instructions. * Capability inventory: Includes the ability to execute git write operations such as git commit, git push, and git worktree prune. * Sanitization: No sanitization or validation of the instructions found in these external sources is mentioned.

git