generate-task
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted product requirement data and transforms it into engineering task documents containing verification instructions for an AI agent.\n
- Ingestion points: The skill reads external requirements content provided via
$ARGUMENTSinSKILL.md.\n - Boundary markers: While the input requirements are delimited by
<prd>tags in the skill context, the generated output task files (TASK_N.md) do not use explicit boundary markers or sanitization to prevent injected instructions from being misinterpreted by the executing agent.\n - Capability inventory: The generated tasks include verification steps that involve executing shell commands such as
tsc,lint,test,build, andcurl.\n - Sanitization: The skill lacks explicit sanitization, filtering, or validation logic to ensure that instructions from the PRD are not maliciously mapped into the execution plan of the generated tasks.
Audit Metadata