emit-pending-decisions

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data and writes it to the local filesystem, creating a vulnerability surface for indirect prompt injection.\n
  • Ingestion points: SKILL.md accepts 'Originating request', 'Evidence', and 'Points' from an invoking caller.\n
  • Boundary markers: The instructions require the agent to quote or summarize the originating request and use structured markdown headers to separate the content.\n
  • Capability inventory: The skill performs file-write operations to create new files within the '.pending-decisions/' directory.\n
  • Sanitization: SKILL.md contains explicit constraints requiring that the suggested follow-up be a natural-language paragraph and 'never an executable command, never machine-readable state, and never a promise that a named skill can be invoked automatically.'
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 08:08 AM
Security Audit — agent-trust-hub — emit-pending-decisions