The Agent Skills Directory

[SAFE]: The skill is assessed as safe. All identified behaviors are consistent with its stated purpose of providing code reviews through static analysis of version control data.
[COMMAND_EXECUTION]: The skill uses standard version control commands, such as git diff and git show, to identify the scope of changes for review. These operations are restricted to the local repository context and are necessary for the skill's primary function.
[DATA_EXFILTRATION]: While the skill reads sensitive source code files, it does not include any instructions or tools for transmitting this data to external servers or non-whitelisted domains.
[PROMPT_INJECTION]: The skill's instructions require the agent to process untrusted data (code changes and PR descriptions), creating a surface for indirect prompt injection. However, the methodology mandates evidence-based findings and structured output, which reduces the likelihood of the agent following malicious instructions embedded in code comments.
Ingestion points: Repository files and diffs accessed via version control tools (SKILL.md).
Boundary markers: None identified.
Capability inventory: Git inspection and file read access; no network-write or code execution tools are requested (SKILL.md).
Sanitization: None identified.

semi-formal-code-review