nd-task-initiation
Pass
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is entirely instructional, providing a framework for the agent to help users who are stuck or avoiding tasks. It focuses on psychological techniques (activation energy, small concrete steps) rather than technical operations.
- [COMMAND_EXECUTION]: The skill contains examples of common shell commands, such as
touch src/models/user.ts. These are presented as literal examples of how to help a user start a project and do not include malicious payloads, obfuscation, or attempts at privilege escalation. - [DATA_EXFILTRATION]: There are no network requests, credential references, or access patterns targeting sensitive system files. The instructions encourage the agent to look at the existing project environment (code, files, configuration) to provide relevant help, which is standard behavior for a development-focused agent.
- [PROMPT_INJECTION]: The skill does not contain instructions to bypass safety guardrails, ignore previous instructions, or extract system prompts. The language used is purely for defining agent behavior in the context of task assistance.
Audit Metadata