dependency-scanning
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill relies on authoritative data from trusted organizations like CISA, FIRST, and Microsoft for vulnerability triage and license categorization.
- [INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for indirect prompt injection but includes strong mitigation measures.
- Ingestion points: Reads manifest files (e.g., package.json, requirements.txt) and lockfiles which are potentially attacker-controlled.
- Boundary markers: Includes a mandatory safety notice that explicitly tells the agent to disregard instructions found in the data.
- Capability inventory: Restricts agent tools to file system reading (Read, Grep, Glob), which limits the impact of any successful injection.
- Sanitization: Enforces a strict output structure and template to ensure the agent's response remains consistent and safe.
- [PROMPT_INJECTION]: The static analysis flag for instruction overrides is a false positive. The strings identified are part of a safety guardrail instructing the agent to resist adversarial inputs.
Audit Metadata