The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local git commands such as status, diff, and commit. It explicitly instructs the agent to perform safe variable quoting when constructing shell commands to ensure that special characters like backticks, dollar signs, and backslashes remain literal, effectively preventing command injection.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of git diff output to generate commit messages. It mitigates potential injection risks through a 'clean-room' drafting policy that requires every line of the message to be traced back to a specific code hunk.
Ingestion points: External data enters through git diff and git diff --cached outputs as specified in SKILL.md.
Boundary markers: None explicitly defined for input, but 'Iron Law 1' and the tracing requirements in Step 2 serve as logical boundaries for content generation.
Capability inventory: The skill utilizes git commit, git hash-object, and git status as identified in SKILL.md.
Sanitization: Step 3 in SKILL.md mandates strict shell variable quoting to sanitize inputs before execution.

commit