orc-quick

Dispatches user-approved subagents to execute small code changes, bug hunts, and PR comment fixes.

6|1|Updated Jul 4, 2026
One-click install
npx skills add https://github.com/azure-id/orc --skill orc-quick-azure-id
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: orc-quick
Source: https://github.com/azure-id/orc/tree/main/templates/skills/orc-quick
Command: npx skills add https://github.com/azure-id/orc --skill orc-quick-azure-id

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Small coding tasks and quick questions often get over-engineered by heavy multi-phase pipelines, or handled without any record of what was decided and why. This Skill provides a lightweight three-step lane — look, ask once, do — that dispatches a user-chosen agent for each task and saves every request as a numbered entry in a persistent context document. ## Core Features & Use Cases - Three-step request handling: Each request runs a silent dig to find relevant files, one combined question turn (clarifications plus agent selection), then a dispatch with build/test checks and a written doc entry. - Mandatory dispatch gate: The user picks the executor, recon model, or reviewer before every dispatch; no config key (including opus5_only) can override or pre-answer this choice. - PR comment workflows: Reads pull request review threads via gh, lets the user pick which threads to fix with per-thread agent gates, and never writes replies, approvals, or merges to GitHub. - Persistent context log: Every request — including read-only investigations — is appended as a numbered entry to orc-quick/<slug>/quick-context.md with decisions, dispatches, files changed, and outcomes. - Use Case: Ask it to "fix the review comments on PR 142" — it fetches the threads, asks which to take and which executor to use per thread, applies the fixes, runs build and tests, and records the outcome before offering to commit. ## Quick Start Ask the skill to quickly find and fix why the orders endpoint returns a 500 error, choosing the executor when prompted.

Frequently Asked Questions about orc-quick

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

FAQPage Schema
How do I fix PR review comments automatically with Claude Code?▼

Invoke the quick lane with the PR number, for example "/orc-quick fix the review comments on PR 142". It reads the review threads via the gh CLI, lets you pick which threads to address, asks which executor agent to dispatch per thread, applies fixes, and offers to commit and push.

What is the difference between orc-quick and the full orc pipeline?▼

The full /orc pipeline runs eight steps including analysis, planning, scoring, and parallel waves. orc-quick runs three steps per request — look, ask once, do — with no spec, plan, or scoring, making it suited to small scoped tasks rather than large features.

Can orc-quick reply to or resolve GitHub review threads?▼

No. It only reads pull request data and pushes code when you explicitly approve. It never runs gh pr comment, never resolves threads, never approves, reviews, or merges — it tells you at ship time that closing threads remains your responsibility.

Why does orc-quick always ask which agent to dispatch?▼

Asking before every dispatch is the lane's core guarantee: you always know what will be spent before it is spent. Config keys like opus5_only are deliberately inert here, and answers are never remembered or reused between requests.

What happens when the build stays red after a code change?▼

A red build triggers a repair loop: rounds one and two reuse your chosen executor, round three asks again so you can pick a stronger model. After three rounds it stops, shows how the error count moved, and lets you choose more rounds, a different executor, or stopping.

When should I not use orc-quick for a task?▼

Avoid it for jobs touching many files, security-sensitive changes, or work needing migration ordering — the lane caps its dig at 12 files and offers to hand oversized jobs to /orc-mini. It also never builds wikis, writes teaching docs, or runs acceptance-criteria verification.