orc-pattern

Codifies a project's real code conventions into cached per-language pattern files for ORC executors.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? AI coding agents often generate code from generic best-practice templates that clash with a project's established house style, producing inconsistent, un-mergeable code. This Skill reconciles generic per-language playbooks against the project's actual most-recently-modified files so generated code matches the existing codebase. ## Core Features & Use Cases - Convention reconciliation: Splits playbook rules into Conventions (project style wins on conflict) and Invariants (security/correctness rules that always apply), then flags real conflicts for the user. - Per-language pattern cache: Writes .claude/orc/patterns/<lang>-pattern.md files covering React, Next.js, Vue, Angular, FastAPI, Django, NestJS, Express, Go, and Postgres, reused by every future ORC run with drift detection and --refresh support. - Subagent-based codification: Dispatches a dedicated codifier subagent to scan representative files, keeping the orchestrator's context lean, with consent-gated scanning and a language-agnostic fallback when declined. - Use Case: Before ORC executes tasks in a FastAPI + React monorepo, run the pattern codifier once so every executor writes endpoints and components that follow the project's actual router layout, naming, and validation style. ## Quick Start Ask the agent to run /orc-pattern to learn and cache this project's code conventions for all detected languages.

Frequently Asked Questions about orc-pattern

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

FAQPage Schema
How do I make AI-generated code match my project's existing style?▼

Run /orc-pattern to codify your project's conventions. It scans the most-recently-modified files per detected language, reconciles them against a generic playbook, and caches the result in .claude/orc/patterns/ for every future ORC run.

Which languages and frameworks does the pattern codifier support?▼

It ships playbooks for React, Next.js, Vue, Angular, FastAPI, Django, NestJS, Express, Go, and Postgres data-access. Postgres is cross-cutting and co-applies with the framework pattern when a task touches the data layer.

What happens when my project style conflicts with best practices?▼

Conventions like folder layout and naming always defer to the project, while security and correctness invariants such as parameterized queries and no leaked secrets are always enforced. Conflicts are flagged to the user once rather than silently resolved.

Does the pattern cache update when my codebase changes?▼

Yes. A cheap structural fingerprint of representative files detects drift, and a drifted cache is automatically re-codified on the next run. You can also force regeneration at any time with /orc-pattern --refresh.

Can I use ORC without running the pattern codification scan?▼

Yes. Set pattern_findings to off or decline at the prompt, and executors fall back to a language-agnostic path: they apply the universal security invariants and imitate the neighboring files they already read, with nothing cached.