agentsclimarketplace

Ultra risk vetting

Skill keithhegit/ultra-orchestration/skills-vnext/ultra-risk-vetting

Ultra Orchestration 是一套开源的 AI 工程编排技能体系。它将 OpenSpec 规格主线（Program → Milestone → Change → Slice）与 Ultra-Orchestrator 执行阶段（Intake → Plan → Dispatch → Execute → Review → QA → Deliver → Retro）结合，提升交付稳定性与可审计性。

Install
npx -y skills add keithhegit/ultra-orchestration --skill ultra-risk-vetting

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Risk classification and approval gating for Codex Ultra vNext. Use before external tools, unfamiliar skills, destructive commands, broad write scopes, credential access, network operations, repository publishing, or any action where safety, reversibility, user approval, or audit evidence matters.

SKILL.md

1.8 KB, as published. Nobody here has run it

Ultra Risk Vetting

Classify risk before execution. This skill is the vNext replacement for the stable risk-vetter skill and absorbs the approval-facing part of safety-guard.

Risk Levels

Use four levels:

  • LOW safe, local, reversible, narrow-scope work
  • MEDIUM normal engineering work with bounded writes or commands
  • HIGH broad writes, destructive operations, publishing, credentials, or external systems
  • EXTREME irreversible, secret-exposing, production-impacting, legal, financial, or security-sensitive actions

Gate Rules

  • LOW: proceed and log briefly
  • MEDIUM: proceed with guardrails and evidence capture
  • HIGH: stop for explicit user approval unless the user already gave concrete approval for the exact action
  • EXTREME: block by default and ask for explicit approval plus scope confirmation

Vetting Checklist

Before execution, inspect:

  • reversibility
  • write scope
  • command destructiveness
  • external network or system effects
  • credential or secret exposure
  • production or user data impact
  • whether the action changes published state
  • whether the action can be safely retried

Output

Return a VetterDecision:

  • risk_level
  • decision
  • required_guardrails
  • approval_required
  • evidence_to_capture
  • reasoning_summary
  • residual_risk

Record the decision in vetter_report and hand required guardrails to ultra-execution-control.

Read risk-policy for examples and thresholds.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.