The Agent Skills Directory

[SAFE]: The skill is a legitimate development tool aimed at facilitating code reviews and technical education within a repository.
[DATA_EXFILTRATION]: The skill contains explicit instructions to redact secrets such as connection strings, tokens, and keys from all generated outputs (reports and chat summaries). It also restricts file writing to specifically designated git-ignored paths (e.g., plans/reports/), minimizing the risk of sensitive context being unintentionally committed to version control.
[PROMPT_INJECTION]: While the skill ingests external data like git diffs and source code, it employs verification protocols requiring that all claims be traced to specific file:line evidence. This reduces the risk of the agent being manipulated by instructions hidden within the analyzed data.
[COMMAND_EXECUTION]: The skill utilizes standard repository tools and local scripts (e.g., git and a python-based code graph utility) for context gathering, which is appropriate and expected for its stated purpose.

understand