The Agent Skills Directory

[SAFE]: The skill is a diagnostic security tool that uses standard repo-scanning patterns. It does not contain any malicious instructions, persistence mechanisms, or unauthorized network operations.
[SAFE]: The skill prioritizes data safety by instructing the agent to never expose raw secret values in output. Instead, it mandates a fingerprinting approach (e.g., providing file path, line number, and prefix only) to prevent accidental credential leakage in chat logs or screenshots.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because its primary function is to read and analyze untrusted external data from code repositories.
Ingestion points: The skill reads repository files, configuration files, and git history using Grep, Read, and Bash git tools as defined in the skill's instructions and reference patterns.
Boundary markers: The skill does not implement specific delimiters for ingested data, but the instructions require the agent to triage every finding into structured categories (Real, False Positive, Needs Review), which acts as a verification step.
Capability inventory: The skill's environment is restricted to read-only repository tools (Grep, git log, git ls-files). It lacks capabilities for writing to the filesystem or making external network requests.
Sanitization: Sanitization is achieved through manual triage where the agent must provide an explicit reason for each finding's classification, ensuring that the processed data is evaluated for context rather than blindly followed.

audit-security