agentleFS
ontology-of-everythingontology-of-everything/SemanticSkills/CLAUDE.md

SemanticSkills

CLAUDE.md14 starsChanged 22 days agoApache-2.0View on GitHub

Concept design, semantic modeling, knowledge organization, visual teaching, and Agent DX tooling. Learning copies, downloaded repositories, process notes, raw outputs and generated eval workspaces belong outside the checkout, normally in a task-specific temporary directory. Delete them when no longer needed. Preserve packaged skills/**/references/ and maintained fixtures/examples. - Update the owning repository; keep sibling projects independent and link rather than copy their skills. - Keep localized pairs, declared source versions and catalog.yml consistent; the catalog is the public version record. English…

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this CLAUDE.md into the rules it gives an agent

12 rules. About 0 are common, 5 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Update the owning repository.

    line 29
  2. Very specific:

    Keep localized pairs, declared source versions and catalog.yml consistent.

    line 30
  3. Very specific:

    Run ./tools/validate-all.sh for public package checks.

    line 34
  4. Very specific:

    Run ./tools/validate-all.sh --local-qa only for available local QA.

    line 35
  5. Very specific:

    Commit and publish only within user authorization.

    line 38
  6. Distinctive:

    Treat credentials, local installs and caches as local-only.

    line 40
  7. Very specific:

    Keep qa/, docs/, analysis material and generated reports local.

    line 42
  8. Distinctive:

    Use these terms consistently in both locales.

    line 50
  9. Very specific:

    Keep supporting material in references/.

    line 83
  10. Distinctive:

    Never report syntax checks as semantic proof.

    line 88
  11. Distinctive:

    Keep decision-record status values and format examples consistent with the checker.

    line 89
  12. Very specific:

    Keep references/diagram-contract.md and scripts/concept_map.py aligned.

    line 92

≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

concept-skills

Concept design, semantic modeling, knowledge organization, visual teaching, and Agent DX tooling.

Repository ownership

RepositoryOwns
concept-skillsSoftware concepts, semantic modeling, knowledge organization, visual learning, and Agent DX tooling.
cloud-concept-skillsHuawei Cloud operations, billing, controlled provisioning, and account onboarding.
handy-skillsOffice workflows, source notes, and Agent DX tooling.

Directory contract

DirectoryContract
skills/en/<name>/, skills/zh-CN/<name>-cn/Self-contained install payloads, including required references/, assets and scripts.
qa/Private local validation, fixtures and reports; ignored by Git and absent from public checkouts.
catalog.ymlTracked discovery metadata used by QA and publishing tools.
docs/Concise local notes and examples; ignored by Git and never required by QA or distribution.
tools/, .githooks/, .github/Repository tooling, local commit/push gates and CI.

Learning copies, downloaded repositories, process notes, raw outputs and generated eval workspaces belong outside the checkout, normally in a task-specific temporary directory. Delete them when no longer needed. Preserve packaged skills/**/references/ and maintained fixtures/examples.

Change and release rules

  • Update the owning repository; keep sibling projects independent and link rather than copy their skills.
  • Keep localized pairs, declared source versions and catalog.yml consistent; the catalog is the public version record. English uses the base name and language: en; Chinese adds -cn and language: zh-CN. Paired translation_of values name each other. Translate prose while preserving commands, safety gates and observable behavior. The repository VERSION and per-skill catalog versions have separate scopes.
  • Run ./tools/validate-all.sh for public package checks; it must work without qa/. Run ./tools/validate-all.sh --local-qa only for available local QA; no behavioral or live-service result is implied by public checks.
  • Set git config core.hooksPath .githooks once per checkout. Ignore rules and commit/push/CI hygiene checks keep local-only files out of Git, including files added with force.
  • Commit and publish only within user authorization. ClawHub publishes the editions enabled in catalog.yml; Chinese skill IDs retain -cn. Verify remote contents before announcing a release.
  • Treat credentials, local installs and caches as local-only. Keep only the current release notes; rewriting Git history or deleting remote releases requires explicit user authorization.
  • Keep qa/, docs/, analysis material and generated reports local; exclude them even when force-added.

Skill authoring structure

Organize each SKILL.md around its scope, inputs, steps, and completion checks. Add a heading only when it separates useful content. Keep definitions and specification syntax in the shared contract; load task-specific references as needed. Keep each sentence to one point. Use the same term for the same concept. Use RFC 2119 keywords for testable obligations, not general advice. Keep the concept entry and the other three capabilities compact.

Domain language

Use these terms consistently in both locales. Preserve syntax, file names, quoted source text, and external standard names.

TermChineseMeaning
model模型The design's concepts, behavior, and composition.
specification规格The written expression of a model.
contract契约Allowed behavior that an implementation must preserve.
specification conventions规格约定This repository's notation and interpretation rules in spec-notation.md.
concept概念An independent software behavior with its own purpose, state, and actions.
concept instance概念实例A particular instance that owns its state.
operational principle, OP运作原理,OPA representative history that explains how a concept fulfills its purpose.
composition组合Application-specific coordination outside concepts, organized into cohesive rule groups.
synchronization rule同步规则External coordination of concept actions. Code uses a <BusinessAction>Synchronization name.
port, adapter端口、适配器An inner contract for needed interaction, and its technical implementation.
facade门面The public application entry; each operation forwards once to a synchronization.
action invocation动作调用A request to execute an action.
action completion动作完成记录A declared result of an executed action.
action occurrence动作发生记录An identified execution that connects invocation and completion.
runtime flow运行时 flowA root event and its causal descendants.
query查询A read that returns binding rows without an action completion.
grain粒度What one fact record represents.
fact, dimension, measure, metric事实、维度、度量、指标Analytical terms defined by the data skill. They are not Jackson concept categories.

The personal-knowledge skill uses concept for a reusable method and states that local meaning explicitly. Do not apply one skill family's qualification criteria to a different family. Keep confirmed, proposed, inferred, and unresolved statements distinct. Reuse existing confirmation when its scope and content still apply.

Concept-specific rules

  • catalog.yml is the discovery source of truth. After catalog changes, run python3 tools/skill-catalog.py --write; the validation gate checks generated README sections and skills.sh.json.
  • Concept Design has one entry per language: concept-design and concept-design-cn. The former eight workflows are combined here; the other seven entries are retired, not aliases. The other three capability families remain separate.
  • Maintain core definitions and principles in concept-design/SKILL.md; keep specification conventions in references/spec-notation.md. Link directly to each modes/<mode>.md. Keep supporting material in references/; name files by their content and responsibility. There are no shared skill dependencies or generated cross-package copies.
  • The modes are prompt selectors, not shell commands. scripts/concept_lint.py is the actual read-only concept-lint CLI. It checks local specification structure, when/then records, and rule-name uniqueness across supplied files for one app. Unknown/legacy call notation returns not_checked/exit 3; only format_checked exits 0. Never report syntax checks as semantic proof.
  • Keep decision-record status values and format examples consistent with the checker. Public checks run packaged examples and representative failing inputs.
  • guard map generates the local Concept and Synchronization Diagram, not Jackson's purpose-dependence diagram. Keep references/diagram-contract.md and scripts/concept_map.py aligned. Version 1 emits one Markdown document; it shares lexical handling with concept-lint and never fills unknown references by inference.
  • Diagrams use Mermaid in Markdown. Keep explanation views concise and use the map generator for concept/synchronization diagrams.
  • Author locale packages directly under skills/ and register them in catalog.yml; keep QA and generated artifacts local.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.