decide

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from board memos and writes it to durable memory locations (~/.claude/decisions/approved/) that are used to ground future agent sessions. This creates an indirect prompt injection surface where malicious content in a memo could influence future agent behavior.
  • Ingestion points: Reads board memo files provided via the argument as described in SKILL.md.
  • Boundary markers: The instructions do not define specific boundary markers or explicit warnings to ignore embedded instructions for the processed content.
  • Capability inventory: The skill performs file system write operations to ~/.claude/decisions/ and ~/company-vault/.
  • Sanitization: The workflow extracts structured data but does not specify sanitization or filtering for free-text fields such as 'Preserved Dissent'.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 05:04 AM
Security Audit — agent-trust-hub — decide