orc-learn

Generates per-feature onboarding documentation with file-anchored walkthroughs and freshness-tracked refresh.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Developers joining a project or touching an unfamiliar feature lack deep, function-level documentation; wikis stay at contract level and go stale. This Skill generates local, per-feature onboarding docs that teach one feature well enough to safely extend it. ## Core Features & Use Cases - INIT mode: Picks one feature (wiki topics first, or a user-pointed directory), then dispatches a pinned writer agent to produce learning-docs/<feature>/ with learning.md (mental model, guided walkthrough, change recipes, FAQ) and knowledge.md (file:line-anchored functions, call flow, contracts, fingerprints). - REFRESH mode: Computes per-feature freshness (FRESH/AGING/STALE) from git distance and drift against stored source_commit and file hashes, then regenerates only the features the user selects. - Use Case: A developer runs /orc-learn, picks the "orders" feature from the wiki list, and receives a pedagogical walkthrough tracing a real request from route to persistence, plus a reference doc with anchored invariants and verify commands. ## Quick Start Ask the assistant to run /orc-learn and pick a feature of this repository to generate local onboarding docs for it.

Frequently Asked Questions about orc-learn

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

FAQPage Schema
How do I generate onboarding documentation for a feature in my repo?▼

Run /orc-learn and pick one feature, either from the wiki's feature areas or by pointing at a directory with a focus hint. The skill dispatches a writer agent that maps the feature's functions, traces one real call flow, and writes learning.md and knowledge.md under learning-docs/<feature>/.

How do I refresh stale generated documentation?▼

Run /orc-learn refresh to list every generated feature with a computed FRESH, AGING, or STALE flag based on git distance and drift since the recorded source_commit. You multi-select which features to regenerate; unselected features are left byte-untouched.

Does orc-learn work without a wiki in the repository?▼

Yes. Without a wiki, or with a stale one, the skill asks you to point at the feature via a directory or focus hint and the writer performs a targeted scan of only those files, never a repo-wide scan.

Are the generated learning docs committed to the repository?▼

No. The learning-docs directory is local and git-ignored; each developer regenerates their own copy. The skill offers to append a learning-docs/ line to .gitignore only after explicit confirmation.

What is the difference between learning.md and knowledge.md?▼

learning.md is the pedagogical half: mental model, guided walkthrough, change recipes, gotchas, and a minimum five-question FAQ. knowledge.md is the reference half: file:line-anchored functions and flow, contracts, extension points, verify commands, and the fingerprint header used for freshness checks.