teach-me

Pass

Audited by Gen Agent Trust Hub on May 4, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an inherent surface for indirect prompt injection as it ingests user-provided topics and answers, stores them in session files (session.md, learner-profile.md), and re-processes them during resumed sessions. Malicious instructions provided as a topic or answer could theoretically influence the agent when the session is loaded.
  • Ingestion points: SKILL.md Step 0 (topic extraction) and Step 3b (tutor responses).
  • Boundary markers: No explicit delimiters (e.g., XML tags) or boundary markers are used to isolate user-stored content from system instructions when re-loading session state.
  • Capability inventory: File system access for reading and writing session data and summaries within the .claude/ workspace.
  • Sanitization: The skill employs a 'kebab-case' slugification process for directory names, providing basic path sanitization for directory creation.
Audit Metadata
Risk Level
SAFE
Analyzed
May 4, 2026, 12:07 PM
Security Audit — agent-trust-hub — teach-me