mindmap
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data and includes self-referential safety claims that target the analysis environment.
- Ingestion points: Processes Plain Text node structures and LLM streaming content as specified in
SKILL.md. Additionally, the filemindmap_audit_result_v2.jsoncontains pre-computed safety verdicts ('security: PASS') designed to influence evaluation. - Boundary markers: The mindmap data format lacks explicit delimiters or boundary markers to separate data from potential instructions within the input.
- Capability inventory: Rendering is performed via local HTML/JS assets (
assets/local-mindmap/index.html), and the skill supports execution of Python scripts under thescripts/directory. - Sanitization: Documentation in
references/streaming-implementation.mdmentions the use of node labeling rules to sanitize and normalize text, providing a baseline mitigation for data ingestion. - [SAFE]: The skill follows security best practices by operating in offline or air-gapped environments and explicitly avoids external JavaScript/CSS libraries, which reduces the risk of supply chain attacks and remote data exfiltration.
Audit Metadata