The Agent Skills Directory

[SAFE]: The skill defines a local development workflow for git operations. It incorporates security-conscious steps, specifically directing the agent to perform sanity checks for secrets, tokens, and debug logs in step 4 of the workflow. All operations are standard local git commands (status, diff, add, commit) without network exfiltration or remote code execution risks.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with the local repository working tree via git commands (SKILL.md) to generate descriptions of changes. This identifies an ingestion surface for potentially untrusted code content; however, the risk is considered safe as the skill's primary function is to assist a developer with commits which are subject to review. Capability inventory is limited to standard git operations and repository-specific tests.

commit-work