expert-panel
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface through untrusted data ingestion.\n
- Ingestion points: User-supplied context is interpolated into the prompt placeholder
【填写问题、已知事实、目标和现实约束】inSKILL.md.\n - Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the interpolated text.\n
- Capability inventory: The skill performs complex reasoning and triggers downstream tool calls to
steelman-both-sidesandminimum-experiment.\n - Sanitization: There is no evidence of input validation or escaping for the user-provided content before it is processed by the agent.\n- [SAFE]: The skill configuration in
agents/openai.yamland instructions inSKILL.mddo not contain any hardcoded credentials, network operations, or executable code. The mention of ascripts/sync.pyfile is for repository maintenance and does not involve insecure execution by the AI agent during runtime.
Audit Metadata