skills/diskd-ai/ccbox/ccbox-insights/Gen Agent Trust Hub

ccbox-insights

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill analyzes historical session logs which introduces an indirect prompt injection attack surface.
  • Ingestion points: Untrusted data enters the agent context through the ccbox history and ccbox skills commands specified in SKILL.md when loading log data.
  • Boundary markers: Absent. There are no explicit boundaries or delimiter instructions guiding the agent to treat the contents of the logs strictly as data rather than instructions.
  • Capability inventory: The skill has file-read capabilities (via the ccbox CLI) and file-generation capabilities (creating markdown reports and proposing snippets to append to AGENTS.md).
  • Sanitization: Absent. The instructions do not define any escaping, filtering, or validation mechanisms for the parsed log data before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:46 AM
Security Audit — agent-trust-hub — ccbox-insights