grimoire-record

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes and persists user-provided project definitions and architectural decisions. While this creates a standard surface for indirect prompt injection, the skill mitigates risk through strict write path limitations and explicit authorization requirements.
  • Ingestion points: Reads from user input and project-specific markdown files in .grimoire/ and adr/ directories.
  • Boundary markers: Instructions do not specify delimiters for user-provided content, but rely on structured markdown templates for recording.
  • Capability inventory: File system read/write access is restricted to .grimoire/CONTEXT.md, CONTEXT-[domain].md, and adr/*.md paths.
  • Sanitization: The skill focuses on factual recording and consistency checks rather than content sanitization.
  • [SAFE]: The skill explicitly forbids critical actions like bootstrapping the knowledge store without proper initialization, writing to unauthorized paths, or bypassing user-initiated authorization scopes. These constraints demonstrate a secure-by-default design for documentation management.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 11:06 AM