recipe-fullstack-implement
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious code, exfiltration patterns, or obfuscation were detected. The skill maintains high safety standards through structured workflows and manual approval steps.
- [PROMPT_INJECTION]: The skill processes user input via the
$ARGUMENTSvariable to route development tasks, creating a surface for indirect instructions. Evidence Chain: (1) Ingestion points: User-provided arguments are processed in the Execution Decision Flow within SKILL.md. (2) Boundary markers: All sub-agent prompts are appended with a [SYSTEM CONSTRAINT] block to scope operational boundaries. (3) Capability inventory: The orchestrator manages agents capable of reading the codebase, creating design documents, and executing code commits. (4) Sanitization: The skill mandates a security-reviewer agent to inspect all implementation outputs for compliance and safety prior to completion.
Audit Metadata