The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external dependency files (e.g., package.json, requirements.txt), which creates a potential surface for indirect prompt injection if those files contain malicious instructions disguised as data.
Ingestion points: Project dependency configuration files accessed via the file-read capability.
Boundary markers: The instructions do not define specific delimiters or clear 'ignore' directives to prevent the agent from potentially following instructions embedded within the audited files.
Capability inventory: The skill is highly restricted, with no network, shell, or file-write capabilities allowed, which significantly limits the impact of any potential injection.
Sanitization: No explicit sanitization or structured parsing requirements are specified for the raw file content before the agent evaluates it.
[METADATA_POISONING]: The skill's YAML frontmatter includes self-assigned safety metrics which could be misleading if relied upon as authoritative.
Evidence: The metadata fields trust-score: 93 and last-audited: '2026-02-03' are claims provided by the skill author about its own security status. As per standard security auditing, these self-reported values should not be treated as a guarantee of safety.

dependency-auditor