dev-pipeline
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided data from the
$ARGUMENTSvariable and propagates it through multiple subagent phases (Analysis, Investigation, Design, Implementation). - Ingestion points: The
$ARGUMENTSinput is written torequest.md, which serves as the primary context for all subsequent subagent invocations. - Boundary markers: The skill lacks explicit boundary markers or instructions for subagents to treat the content of
request.mdas potentially untrusted data that should not be interpreted as commands. - Capability inventory: Downstream subagents, particularly in the Implementation and Final Verification phases, have the capability to write source code, create files, and execute shell commands (e.g.,
git, test suites, and linters). - Sanitization: There is no evidence of sanitization or validation of the input before it is interpolated into prompts or written to the workspace artifact files.
Audit Metadata