skills/shinpr/nautilus/recipe-reflect/Gen Agent Trust Hub

recipe-reflect

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes data from multiple source files and user input to update product documentation, creating a surface for indirect prompt injection.
  • Ingestion points: Data enters the agent context through $ARGUMENTS (SKILL.md), hypothesis files (Level 1 reflection), and Opportunity files (Level 2 reflection).
  • Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the ingested hypothesis or opportunity data.
  • Capability inventory: The skill performs file-write operations to several locations, including docs/product/vision.md, docs/product/learnings.md, docs/discovery/INDEX.md, and various hypothesis/opportunity files.
  • Sanitization: No explicit sanitization, validation, or filtering of the external content is performed before interpolation into the prompt for the sub-agent or before writing to target files.
  • Mitigation: The workflow requires a [STOP — BLOCKING] checkpoint where the user must explicitly confirm findings from the knowledge-distiller before any learnings are promoted or written to the knowledge pyramid.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 10:15 PM
Security Audit — agent-trust-hub — recipe-reflect