The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git commands for staging and committing, and runs project-specific validation tools such as npm test or custom verification scripts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of code diffs, which could potentially contain adversarial instructions intended to influence the agent's behavior during the commit process.
Ingestion points: git diff and git diff --cached output processed in SKILL.md.
Boundary markers: None used to isolate untrusted diff content from the skill's instructions.
Capability inventory: Includes git commit operations and execution of arbitrary shell commands for testing.
Sanitization: No explicit sanitization or validation of diff content is performed.
[EXTERNAL_DOWNLOADS]: The documentation references external repositories for its origin and integrated components.
Evidence: References to the anthropics/skills repository and the nesnilnehc/ai-cortex project for provenance and integrated pre-commit reviews.
[SAFE]: The skill adopts security-positive practices by instructing the agent to actively check for sensitive data (secrets, tokens, or debug logs) before finalizing any commits.

commit-work