The Agent Skills Directory

[SAFE]: The skill instructions and logic are consistent with its stated purpose of codebase analysis. No patterns associated with obfuscation, data exfiltration, or credential theft were found.
[COMMAND_EXECUTION]: The skill incorporates standard shell commands (git log, wc, sort, uniq, head) to extract metadata from the repository history. These commands are non-destructive, read-only, and typical for software development workflows.
[PROMPT_INJECTION]: The skill operates on external data (source code and git history), which constitutes an indirect prompt injection surface. The following evidence chain is noted for this surface:
Ingestion points: The agent is instructed to read various files including README.md, entry points, manifests, and the source files within the repository.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions embedded in the analyzed code.
Capability inventory: The skill utilizes file-reading capabilities and local subprocess execution for git commands. It does not define network operations or file-writing tasks.
Sanitization: No explicit sanitization or filtering of the ingested content is described. While the surface exists, the risk is categorized as safe because the skill does not expose dangerous capabilities that could be abused via injected instructions.

code-analysis