dbs-learning
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from local project files and user feedback sections to calibrate its teaching gradient. It mitigates potential risks through explicit boundary markers and source verification.
- Ingestion points: The skill reads root-level project files (e.g.,
README.md,CLAUDE.md) and user-modified feedback areas within learning articles, as specified in Phase 1.5 and Phase 2 ofSKILL.md. - Boundary markers: It employs specific markdown headers (
## 学习反馈) and template filtering instructions to isolate user input from instructional logic. - Capability inventory: The skill's actions are restricted to reading project files and writing markdown content within authorized directories; it does not leverage network or shell execution capabilities.
- Sanitization: It includes an explicit exclusion list for sensitive paths (e.g.,
.env,.git/, keys, secrets) and requires the agent to differentiate between user-verified content and AI-generated drafts to prevent context contamination.
Audit Metadata