skills/a-canary/arc-skills/counsel/Gen Agent Trust Hub

counsel

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from a user-provided topic and the content of "relevant files" read during the investigation phase.
  • Ingestion points: The parameter in ke:recall and the content of files accessed via "read relevant files" as described in the Workflow section of SKILL.md.
  • Boundary markers: Absent. There are no instructions or delimiters (such as XML tags or "ignore instructions" headers) defined for the sub-sessions when processing external data.
  • Capability inventory: The expert sub-sessions have the capability to perform file reads, memory retrieval (ke:recall), and generate actionable reports that influence the final synthesis.
  • Sanitization: Absent. The skill does not specify any validation, escaping, or filtering on the content of the files or the topic string before presenting them to the expert models.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 08:52 PM
Security Audit — agent-trust-hub — counsel