go
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary user goals via
$ARGUMENTSto generate specifications and execution plans for autonomous agents. A malicious goal could influence the generated logic or instructions executed by the conductor-orchestrator. - Ingestion points:
SKILL.md(via$ARGUMENTSin the/gocommand). - Boundary markers: Absent; user input is not delimited or warned against containing embedded instructions.
- Capability inventory: File system access (creating tracks,
spec.md,plan.md, andmetadata.json), directory creation, and delegation to theconductor-orchestratoragent. - Sanitization: Absent; no input validation, escaping, or filtering is performed on the user-provided goal before it is used to generate the plan and metadata.
Audit Metadata