sop

Orchestrates multi-agent coding workflows with human approval gates and capability-based roster assignment.

69|1|Updated Jul 17, 2026
One-click install
npx skills add https://github.com/ythx-101/agent-sop --skill sop-ythx-101
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: sop
Source: https://github.com/ythx-101/agent-sop
Command: npx skills add https://github.com/ythx-101/agent-sop --skill sop-ythx-101

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Coordinating multiple coding agents on a task often lacks structure: agents self-approve risky actions, reviewers rubber-stamp their own work, and no audit trail exists for deployments. This Skill provides a standard operating procedure that assigns agents to capability slots, enforces human approval gates, and produces signoff documentation for every high-risk change. ## Core Features & Use Cases - Roster Discovery: Runs scripts/discover-roster.py to detect installed agent CLIs, probe their health, and assign slots (sole-writer, reviewer, scout, monitor) by capability tier and vendor independence rather than hardcoded names. - Tiered Workflow Weight: Classifies tasks as read-only, medium, or data/service/destructive, applying proportionate process from direct execution to the full research-plan-approve-implement-review-verify-signoff-deploy pipeline. - Independent Review & Signoff: Enforces cross-vendor read-only review with a VERDICT/BLOCKING/SUGGESTED/EVIDENCE contract, plus Markdown signoff packets where only humans can set approved status. - Use Case: A team running Claude, Codex, and Gemini CLIs on a shared host uses this Skill so a database migration gets a written plan, explicit human approval, sole-writer implementation, cross-vendor review, and a signed deployment record. ## Quick Start Ask your agent to use the sop skill to plan and implement your task with independent review and human signoff before any deployment.

Frequently Asked Questions about sop

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I coordinate multiple AI coding agents on one task?

Use a capability-slot workflow: an orchestrator discovers available agent CLIs, assigns exactly one sole-writer per repository, and dispatches read-only reviewers and scouts. Human approval gates sit between planning, implementation, and deployment phases.

How to detect which coding agent CLIs are installed and healthy?

Run scripts/discover-roster.py with the --probe flag. It scans PATH for known agent CLIs, checks versions, and sends a minimal non-interactive prompt to each, catching expired auth or provider outages before work is assigned.

Does this workflow work with agents from different vendors?

Yes, cross-vendor review is a core requirement. The reviewer slot prefers the highest-capability healthy agent from a different vendor than the writer, with a documented degradation ladder when full independence is unavailable.

Can AI agents deploy or publish code without human approval?

No. Gated actions like deploys, restarts, data writes, and publishing are never self-served by agents. The orchestrator splits work at each gate, and signoff status can only be set to approved by a human, quoted verbatim.

When should I use the full workflow versus the lightweight path?

Read-only or tiny reversible tasks run directly with an after-action report. Medium tasks need a plan, approval, and independent review. Any production data, service, or destructive change requires the full pipeline including a signoff packet.