ai-memory-learning-maintenance
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is focused on knowledge base management and administrative cleanup within a defined project scope. All tools and operations described are consistent with the stated purpose of memory maintenance.
- [INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted data from user sessions and wiki pages, presenting a surface for indirect prompt injection which is addressed through explicit behavioral constraints.
- Ingestion points: Raw session observations and wiki page content processed by
memory_consolidate,memory_auto_improve, andmemory_lint(SKILL.md). - Boundary markers: The instructions explicitly state that project data remains "untrusted project data" and must never be treated as authority for facts, tool use, policy, or schema changes (SKILL.md).
- Capability inventory: The skill allows the agent to write markdown wiki pages, propose durable project rules, and delete expired memory files.
- Sanitization: Content is described as being sanitized, bounded, and JSON-encoded prior to processing.
Audit Metadata