The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several git commands (git status, git diff, git branch, git add, git commit) and development tools via the bun runtime (bun run lint, bun run test, bun run format). These operations are necessary for the skill's stated purpose of managing code commits and ensuring quality.
[PROMPT_INJECTION]: The skill ingests untrusted data by reading local code changes through git diff. This introduces a surface for indirect prompt injection if malicious instructions are embedded within the code files being analyzed.
Ingestion points: Untrusted data enters the agent context through git diff in Step 1 of SKILL.md.
Boundary markers: No explicit boundary markers or delimiters are used to separate the code content from the agent's instructions.
Capability inventory: The skill has the capability to perform file staging, commits, and execute local build/test scripts via bun.
Sanitization: There is no automated sanitization of the input; however, the skill mandates a human-in-the-loop check by requiring the user to approve commit messages before execution.

commit