The Agent Skills Directory

[SAFE]: The skill implements a professional software engineering methodology for codebase audits, focusing on structural health and maintainability.
[SAFE]: The execution environment is appropriately restricted via the allowed-tools frontmatter, limiting shell access to specific git commands and user-defined verification scripts.
[SAFE]: The skill does not contain any suspicious patterns related to data exfiltration, credential theft, or unauthorized persistence.
[SAFE]: No obfuscation techniques or hidden instructions were found in the skill's code or documentation.
[SAFE]: The skill ingests untrusted code for analysis, which is a necessary operational surface for its stated purpose.
Ingestion points: Phase 2 instructions specify reading "every file in the source directory" using Read, Grep, and Glob tools.
Boundary markers: None present; the agent processes raw file content directly to evaluate code quality.
Capability inventory: Subprocess access is limited to the Bash tool for git operations and designated check commands.
Sanitization: No sanitization is performed on analyzed code, as the agent must evaluate the original source for accurate reporting.

codebase-review