The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill invokes npx -y gitzy commit, which downloads and executes the gitzy package from the npm registry. This package is maintained by the skill's author and is used to generate conventional commit messages. This pattern involves running remote code at runtime to perform the primary function of the skill.\n- [EXTERNAL_DOWNLOADS]: The skill fetches the gitzy tool from the npm registry at runtime via npx. While common for development tools, this introduces an external dependency on the integrity of the remote package and registry.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted data from the repository's git history and current changes to infer commit conventions and draft messages, presenting a potential surface for injection.\n
Ingestion points: Untrusted data enters via repository history (Step 1) and current file changes (Step 2).\n
Boundary markers: No specific delimiters or markers are defined to isolate untrusted content from the surrounding prompt instructions.\n
Capability inventory: The skill can perform file system operations (git add/commit) and execute a local Python script for linting.\n
Sanitization: The skill lacks explicit sanitization or filtering of the ingested git data before it is used for inference or command construction.\n- [COMMAND_EXECUTION]: The skill performs multiple shell operations, including git commands for staging and logging, and calls a local Python script (scripts/lint_commit.py) using python3. The test suite similarly uses subprocess.run to call the linter script in an isolated Python environment.

commit