license-compliance-triage

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data such as dependency lists, SBOMs (Software Bill of Materials), and license reports which are often sourced from third-party repositories or generated by automated tools. Malicious actors could embed prompt injection strings within dependency metadata or license expressions (e.g., a package name or license field containing instructions to the LLM).
  • Ingestion points: The skill explicitly processes THIRD-PARTY.txt, SBOMs, Maven/npm dependency manifests, and license reports provided by the user (as described in SKILL.md).
  • Boundary markers: The instructions lack explicit requirements for the agent to wrap untrusted input in strict delimiters or to ignore instructions embedded within the data fields.
  • Capability inventory: The agent is tasked with interpreting complex expressions and generating actionable compliance reports, which involves deep processing of the untrusted strings.
  • Sanitization: While the skill provides instructions to desensitize credentials (tokens, private URLs), it does not provide specific sanitization or filtering for instructional content embedded in the data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 03:49 AM
Security Audit — agent-trust-hub — license-compliance-triage