osv-vulnerability-scan
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch vulnerability information from the official OSV API (
api.osv.dev) and references documentation hosted on Google's GitHub pages (google.github.io). These are well-known and trusted services for security analysis. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from package manifests, lockfiles, or user lists to generate vulnerability queries.
- Ingestion points: Workflow step 1 in
SKILL.mdidentifies packages from external manifests or user input. - Boundary markers: The instructions do not specify explicit delimiters for interpolating external package data into the API request bodies.
- Capability inventory: The skill utilizes
curland PowerShellInvoke-RestMethodto execute network queries based on the ingested package data. - Sanitization: The skill includes specific validation rules in
SKILL.md("Query rules (avoid 400)") to ensure API compatibility, though it does not specify content sanitization for the inputs themselves. - [SAFE]: The skill includes proactive safety instructions, such as prohibiting the creation of exploit PoCs and explicitly warning against the exposure of secrets from environment files or private registries.
Audit Metadata