Name the meaning first — then write code, run a CLI, or draft a spec.
concept-skills provides nine
capabilities as 18 localized Agent Skills for ontology, semantic layers,
concept design, repository guidance, and presentation authoring. English uses the base name;
Simplified Chinese adds -cn.
Huawei Cloud operational skills now live in
concept-git/cloud-concept-skills and are no
longer distributed from this repository.
The concept-design skills adapt Daniel Jackson's concepts-and-synchronizations model — The Essence of Software (2021), with the current when/where/then sync notation from Beyond Objects (arXiv:2606.27258) — for agent use; an adaptation, not endorsed by the author.
中文说明见 README-CN.md.
Current release is 1.1.0. The concept-* family is explicit-only; concept-guardrails modes
consume Jackson notation only.
Requirements to modules, with the concept model as the contract. Each skill stops at its own
boundary and hands off: design → prd / implementation → audit.
| English | 中文 | What it does |
|---|---|---|
concept-design |
concept-design-cn |
Designs independent concepts and synchronizations. |
concept-prd |
concept-prd-cn |
Transcribes confirmed models into traceable PRDs and specifications. |
concept-implementation |
concept-implementation-cn |
Implements independent modules and synchronization composition. |
concept-audit |
concept-audit-cn |
Audits design arguments, drift, composition, and optional code conformance. |
concept-guardrails |
concept-guardrails-cn |
Creates, audits, compares, and maps colocated specifications. |
Enterprise knowledge from APIs and databases (semantic-km-creator); personal knowledge from source
text (semantic-pkm-creator).
| English | 中文 | What it does |
|---|---|---|
semantic-km-creator |
semantic-km-creator-cn |
Builds evidence-only Kimball semantic layers from interfaces. |
semantic-pkm-creator |
semantic-pkm-creator-cn |
Extracts scenes, concepts, and entities from source text. |
| English | 中文 | What it does |
|---|---|---|
html-slides |
html-slides-cn |
Builds single-file styled HTML decks with optional PPTX. |
claude-md |
claude-md-cn |
Creates or prunes CLAUDE.md from repository evidence. |
Per-skill details: docs/skills/en/ and docs/skills/cn/. Machine-readable index:
docs/catalog.yml. Localization contract:
docs/localization.md.
Requires Node.js for npx.
Base names install English. Append -cn for Simplified Chinese.
ClawHub publishes only the English skills/en editions; Chinese editions remain on GitHub.
npx skills add ontology-of-everything/concept-skills \
--skill <skill-name> \
--agent cursor \
--copy -y--agent accepts cursor, claude-code, or codex. Several skills at once, and --global to
install into ~/.agents/skills/ (scanned by both Cursor and Codex) instead of the project's
.agents/skills/:
npx skills add ontology-of-everything/concept-skills \
--skill concept-design concept-prd concept-guardrails \
--agent cursor codex \
--global --copy -yList what's available, or install from a local checkout while developing:
npx skills add ontology-of-everything/concept-skills --list
npx skills add ./skills/<skill-name> --skill <skill-name> --agent cursor --copy -yDiscovery: skills.sh (groups in
skills.sh.json) · SkillsMP (GitHub topics
claude-skills, claude-code-skill) · ClawHub.
Agent-specific notes: Cursor · Claude Code · Codex.
Read a skill before using it — skills run with your agent's permissions.
Skills normally activate from their description, so plain requests are enough. The concept-*
family is explicit-only and must be named (/concept-design in Cursor, $concept-design in Codex):
$concept-design model this requirement as independent concepts
$concept-prd transcribe the confirmed model into a PRD
$concept-guardrails audit specification drift in src/orders
Turn this API into a semantic layer → semantic-km-creator
Build a 15-minute product presentation → $html-slides
To pin a skill explicitly: /skill-name in Cursor, $skill-name in Codex.
docs/CONTRIBUTING.md · docs/authoring.md
./tools/skill-scaffold.sh <skill-name> # new skill
./tools/install-git-hooks.sh # pre-commit → validate-all.sh
./tools/validate-all.sh # all skills, same as CI
./qa/<skill-name>/validate.sh # one skillskills/en/<name>/ and skills/cn/<name>-cn/ are self-contained install payloads. QA follows the
same locale split and is never installed. Update both locales, catalog, docs, and parity checks
together.
Apache-2.0 © concept-skills contributors. Bundles published to ClawHub are MIT-0 there; the repository source stays Apache-2.0.