orc-wiki

Scans a codebase and builds a persistent, evidence-anchored project knowledge base in wiki/ docs.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? AI coding sessions lose project context between runs, forcing models to re-read code or guess at contracts. This Skill builds and maintains a persistent, evidence-anchored knowledge base under the project-root wiki/ folder so future ORC runs consult verified docs instead of model priors. ## Core Features & Use Cases - Codebase scanning to docs: Dispatches pinned scanner agents (Opus 4.8 high) to write wiki/orc-feature-, wiki/orc-reference-, and orc-architecture-overview.md docs whose claims are anchored to real files, with freshness computed on read from git commit distance. - Deterministic registration: Runs the orc wiki sync CLI to derive wiki/INDEX.md and .claude/orc/wiki-meta.json from doc headers, so a paused multi-session scan still leaves a valid, indexed wiki. - Cross-repo crosslink: Publishes per-endpoint boundary tags and resolves consumed contracts from linked sibling repos, with advisory drift warnings and a federation atlas. - Use Case: After a large feature ships, run a targeted refresh so only the docs whose covered files changed are re-scanned, keeping the knowledge base current without paying for a full re-scan. ## Quick Start Ask the assistant to run /orc-wiki to scan this codebase and build the project knowledge base, approving the cost warning when prompted.

Frequently Asked Questions about orc-wiki

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

FAQPage Schema
How do I build a project knowledge base for Claude Code?▼

Run /orc-wiki in the project root. It scans the codebase with pinned scanner agents and writes evidence-anchored docs under wiki/, then registers them with orc wiki sync so future runs can consult them.

How do I refresh stale wiki docs without a full re-scan?▼

Use /orc-wiki update <doc> or /orc-wiki refresh for a targeted refresh. The orc wiki plan command ranks pending work and free repairs like orc wiki sync always run before any paid re-scan.

Why does ORC say it cannot see my wiki?▼

The wiki is likely UNREGISTERED: docs exist but no manifest indexes them, usually because a scan paused before its assemble phase. Run orc wiki sync to rebuild the index instantly instead of re-scanning.

Can orc-wiki share API contracts across multiple repos?▼

Yes. The crosslink subsystem publishes per-endpoint tag files for each repo's boundary and lets a consuming repo resolve real contracts from linked sibling repos, with advisory drift warnings that never block a build.

How is wiki freshness computed?▼

Freshness is computed on read, never stored: git commit distance between each doc's scanned_commit and HEAD, scoped to the files that doc covers. Tiers are FRESH, AGING, or STALE, and code always outranks the wiki on conflict.

Why does /orc-wiki ask for consent before scanning?▼

A full scan is expensive and often multi-session because it uses high-tier models for both orchestration and scanning. The skill never touches the repo until the user explicitly approves the cost warning.