memory-comark
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the aggregation and display of data from the knowledge base through the 'query' attribute in components like 'bm-table' and 'bm-stat'. This creates an indirect prompt injection surface where malicious instructions in queried notes could be presented to the agent.
- Ingestion points: Data ingestion occurs through 'notes', 'search', and 'activity' sources as described in SKILL.md and references/components.md.
- Boundary markers: Visual components are delimited using '::bm-name' and '::' markers.
- Capability inventory: The skill instructions involve the use of 'read_note', 'write_note', 'edit_note', and 'search_notes' tools to interact with the environment.
- Sanitization: The skill mandates that note content avoid HTML, scripts, and styles. Media components like 'bm-image' and 'bm-svg' are restricted to project-local files, preventing remote content inclusion. The query pipeline is restricted to a specific, validated grammar.
Audit Metadata