spec-loop-implementation-flow
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFENO_CODEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is composed strictly of Markdown documentation defining an implementation workflow and contains no executable components.
- [NO_CODE]: No scripts, binaries, or tool configurations are provided. The skill instructs the agent using its pre-existing capabilities.
- [PROMPT_INJECTION]: The skill contains no direct malicious instructions or safety bypasses. It defines a surface for indirect prompt injection through task processing.
- Ingestion points: The agent reads canonical task files and chat history as the source of truth for implementation.
- Boundary markers: No specific delimiters are mandated in the instructions to isolate task data from agent instructions.
- Capability inventory: The skill lacks any capability to execute shell commands, make network requests, or perform dynamic code execution.
- Sanitization: The workflow does not specify any data sanitization or filtering steps.
Audit Metadata