concept-coverage

Warn

Audited by Gen Agent Trust Hub on Aug 25, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTIONNO_CODE
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill implements a self-improvement protocol directing the agent to modify its own source file (SKILL.md) by appending to the 'Things to Remember' section. This self-modifying code pattern is a medium-risk finding as it creates a persistence vector for malicious instructions to be incorporated into the skill's permanent instructions.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and summarizes user-provided content during tracking sessions. 1. Ingestion points: User prompts analyzed for conceptual depth. 2. Boundary markers: Absent. 3. Capability inventory: Local file writing (concept-coverage-log.json and SKILL.md). 4. Sanitization: Relies on the agent to summarize and filter content, which lacks robust validation against adversarial input.
  • [DATA_EXPOSURE]: The skill logs summaries of user interactions and identified knowledge gaps to a local file. There is a risk that sensitive information provided during the session could be captured in these summaries.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 25, 2026, 09:15 PM
Security Audit — agent-trust-hub — concept-coverage