pentest-blackboard
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, including user messages and tool outputs (e.g., from terminal curl and browser_navigate).
- Ingestion points: External data enters through user messages, HITL (Human-in-the-loop) approvals, and outputs from the component-vuln-intel command set (curl/browser).
- Boundary markers: None identified for delimiting external content within the blackboard or facts.
- Capability inventory: The skill uses tools for writing to a database (upsert_project_fact), recording vulnerabilities (record_vulnerability), and performing network operations (terminal curl, browser_navigate).
- Sanitization: The skill mentions desensitizing credentials (脱敏) before storage but lacks explicit sanitization or validation rules for all other external inputs processed into facts.
- [COMMAND_EXECUTION]: The skill instructs the agent to perform network-based scanning using terminal curl and browser_navigate as part of its mandatory 'Behavior Triggers' when a component is identified.
- [CREDENTIALS_UNSAFE]: The skill is designed to capture, store, and reuse credentials (e.g., auth/cred-* facts and 'loot' management). While it mentions desensitization, it explicitly manages authentication artifacts as part of its core penetration testing functionality.
Audit Metadata