memorize

Pass

Audited by Gen Agent Trust Hub on Apr 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from conversation history and external tool outputs to update persistent context.
  • Ingestion points: Conversation history, reflection outputs from /reflexion:reflect, and critique findings from /reflexion:critique (SKILL.md, Phase 1).
  • Boundary markers: The instructions lack explicit delimiters or markers to isolate the harvested data from the agent's logic during the extraction and curation phases.
  • Capability inventory: The skill has the capability to modify the filesystem by writing to CLAUDE.md (SKILL.md, Phase 3).
  • Sanitization: The skill implements manual sanitization rules, such as a prohibition on secrets, tokens, and PII, alongside 'Quality Gates' to review coherence and actionability. However, these checks are performed by the model itself rather than a hard boundary system.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 23, 2026, 03:49 AM