tm-security-review

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze untrusted external data, including whole code repositories, diffs, and existing threat models. This creates a surface for indirect prompt injection where an attacker could embed malicious instructions within the source code or documentation to manipulate the agent's findings or execution.
  • Ingestion points: Target repository path, service features, diffs, and existing *-threat-model.md files specified in the 'Inputs' section.
  • Boundary markers: The skill requires the agent to ask for 'Deployment/trust context' to calibrate severity, but does not explicitly define sanitization or boundary markers to isolate instructions within the ingested data from the agent's core logic.
  • Capability inventory: The skill uses file reading, grep, graph tracing, and manual code review to identify vulnerabilities and generates runnable Proof-of-Concept (PoC) code snippets (curl, payloads, unit tests).
  • Sanitization: No explicit sanitization or filtering of external content is described prior to interpolation into the analysis phases.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 03:58 AM
Security Audit — agent-trust-hub — tm-security-review