ethical-sourcing

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from the local codebase and system configurations without explicit boundary markers or instructions to ignore embedded commands.
  • Ingestion points: Codebase files, system configuration data, and supplier data models.
  • Boundary markers: Absent.
  • Capability inventory: File read access and file write access (specifically to the docs/ and ~/.claude/projects/ directories).
  • Sanitization: No sanitization, escaping, or validation of the ingested content is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:56 AM
Security Audit — agent-trust-hub — ethical-sourcing