typmem-learn
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data—such as user dialogue or existing retro files—to extract and save durable facts. This creates a surface for indirect prompt injection where malicious content could be stored in the local file system. However, the risk is negligible as the skill's capabilities are limited to writing markdown files to a specific local directory (
~/.typmem/memory/learnings/) and it does not trigger any high-privilege actions or network exfiltration. - Ingestion points: User input (commands like "remember this") and existing retro files located in
~/.typmem/memory/retro/. - Boundary markers: None explicitly defined in the instructions for the content of the markdown files.
- Capability inventory: Directory creation (
mkdir) and file writing (~/.typmem/memory/learnings/). - Sanitization: The skill instructs the agent to convert facts into a kebab-case slug for the filename and to distill content into a specific recipe-card format, which acts as a manual form of sanitization.
Audit Metadata