02-vulnerability-scanner
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires standard Python libraries (requests, packaging, jinja2, pyyaml) which are typical for its vulnerability scanning and reporting functionality.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, such as server configuration files, dependency manifests, and network scan results. This introduces a surface for indirect prompt injection attacks where malicious content in these files could attempt to influence the agent's behavior.
- Ingestion points: scripts/dependency_auditor.py and scripts/config_auditor.py read files from the project directory.
- Boundary markers: The skill includes an 'Evidence contract' section in SKILL.md instructing the agent to treat file contents and tool responses as untrusted evidence, separating observed and inferred conclusions.
- Capability inventory: The skill utilizes terminal access to run local scripts, nmap, and other security tools.
- Sanitization: The scripts parse specific data structures but do not implement extensive sanitization of the raw content before agent analysis.
- [DATA_EXFILTRATION]: The dependency_auditor.py script fetches vulnerability advisories from the well-known OSV database (api.osv.dev). This is a standard industry practice for identifying known security flaws in software packages.
Audit Metadata