linddun
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project files and dispatches their content or metadata to subagents, creating an attack surface for indirect prompt injection.
- Ingestion points: The skill ingests a scoped list of project files (
FILE_LIST) provided by the user or environment inSKILL.md. - Boundary markers: The subagent prompt template in
SKILL.mddoes not use delimiters or clear separators to isolate the untrusted file content from the subagent's instructions. - Capability inventory: The skill dispatches
general-purposetask calls to subagents, which perform further analysis based on the potentially poisoned input. - Sanitization: No evidence of sanitization, escaping, or instructions to ignore embedded commands is present in the dispatcher prompt template.
Audit Metadata