`catalogue-curation` — guides

Choose the operator workflow for authoring, compiling knowledge, surveying, assimilating, governing, or redistributing catalogue primitives.

The operator’s toolkit for growing and maintaining an agent-skill catalogue: write new skills to standard, bring good work in from outside (shaped to our craft), compile pack-owned knowledge, survey a whole repo, and derive an owned catalogue for another org or domain. Domain-agnostic — the same toolkit serves a non-SDLC catalogue.

Step 1 — Standards (read this first)

Skill standards — one page. The three standards every skill is measured against, whether you write it or bring it in. Gives you the reading order for what follows.

Step 2 — Choose your path

Main journey: writing a new skill

Building a skill from scratch — frontmatter to PR.

Skill standards — what you’ll be measured against (Step 1 above)
Your first skill — the end-to-end tutorial: workspace design, standards inline at each step, definition of done
How to author a skill — the deep reference once you’ve done the tutorial

Side journey: assimilating an existing skill or repo

Bringing external work into the catalogue — safely, and shaped to convention.

Skill standards — know what you’re measuring incoming material against (Step 1 above)
Choose the entry point for your material:
- Your first assimilation — one skill from a URL
- Your first subagent — one subagent definition; where the OWASP AST review shifts focus
- Survey a repo — a whole source at once, resumably
The convergence model — why assimilation shapes rather than pastes; the three convergence layers in full

Side journey: shipping pack-owned knowledge

Turn a maintained reference corpus into a portable, discoverable router Skill.

Author and compile an OKF bundle — declare a reference-only OKF 0.2 bundle, compile its managed output, check drift, and inspect catalogue discovery
Catalogue format — the surrounding manifest, schema, ownership, and validation contract

Step 3 — Go deeper (when you’re ready)

The convergence model — three layers, two primitive types, the no-merge-back principle, and the pack arc
Catalogue operator journey — how org-stack engineers, catalogue maintainers, and catalogue authors use the pack at their altitude
Why curation is its own pack — the single-authoritative-source model and the fail-closed posture

Full reference

Tutorials

Your first skill — write a new skill from scratch; workspace design, standards inline, definition of done
Your first assimilation — bring one external skill in, safely and to convention
Your first subagent — bring in an external subagent definition; where the OWASP AST review shifts focus

How-to

Author and compile an OKF bundle — project pack-local OKF knowledge into a portable router Skill and verify it is current
Survey a repo for what to adopt — inventory a whole source into a reviewable RFC, resumably
Create a self-hosted catalogue — derive, brand, validate, and package an owned catalogue for another org or domain

Reference

The ledger and the engine guard — where assimilation state lives, and what the guard blocks
Direct-install diagnostic codes — every refusal the direct skill-installation route can emit, and what to do about it
Catalogue CI contract — publication ordering, exit codes, and responsibility boundaries for CI-driven catalogue releases

Explanation

Skill standards — the three standards, the definition of done for both paths, and the reading order
The convergence model — the three convergence layers (agentskills.io, catalogue craft, OWASP AST), two primitive types, and the no-merge-back principle
Catalogue operator journey — how org-stack engineers, catalogue maintainers, and catalogue authors use the pack at their altitude
Why curation is its own pack — the single-authoritative-source model and the fail-closed posture

Installing and upgrading live in ../_shared/.