cw-panel

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text drafts and passes them to multiple subagents and a synthesizer for analysis. This creates a surface where an attacker could embed malicious instructions within a draft to manipulate the reviewer agents' outputs or the final synthesis.\n
  • Ingestion points: The skill reads external drafts and project metadata in Step 1 (SKILL.md) to provide context for the review.\n
  • Boundary markers: There are no explicit instructions or delimiters used when passing the draft content to subagents to distinguish between legitimate content and potential embedded instructions.\n
  • Capability inventory: The skill has the capability to spawn several subagents (compound-writing:review:*) and a synthesizer agent. It does not appear to have direct file-write or network access capabilities.\n
  • Sanitization: No sanitization, validation, or escaping is performed on the ingested draft content before it is processed by the AI agents.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 06:43 PM
Security Audit — agent-trust-hub — cw-panel