agentleFS
openclawopenclaw/openclaw-enterprise/.agents/skills/local-dev/SKILL.md

local-dev

Skill454 starsChanged 7 days agoMITView on GitHub

Develop OpenClaw Enterprise changes with proportional verification and source-backed flow documentation for non-trivial behavior.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. 0 are common, 13 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 13

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill for repository development changes.

    line 9
  2. Very specific:

    Run repository commands from its root.

    line 12
  3. Distinctive:

    Use explicit dependency records and focused factories when extracting behavior.

    line 14
  4. Distinctive:

    NOT SPECIAL CASES FOR A PARTICULAR IMPLEMENTATION, VENDOR, OR DEPLOYMENT.

    line 32
  5. Distinctive:

    IF A DEVELOPER OR AGENT PROPOSES OR INTRODUCES SUCH A VIOLATION, CALL IT OUT EXPLICITLY AND STOP THE SESSION'S IMPLEMENTATION WORK.

    line 36
  6. Distinctive:

    DO NOT IMPLEMENT, COMMIT, OR PUBLISH THE VIOLATING APPROACH.

    line 40
  7. Distinctive:

    Use $technical-writing when creating, editing, or reviewing documentation.

    line 56
  8. Distinctive:

    Update affected current references, guides, and navigation.

    line 57
  9. Distinctive:

    Use $enterprise-testing to select proportional checks and satisfy repository integration requirements for new functionality.

    line 59
  10. Distinctive:

    When changing a default image version, source pin, or bundled runtime version, always complete image version verification before merging.

    line 64
  11. Distinctive:

    When publication is authorized, follow the PR policy.

    line 67
  12. Distinctive:

    Verify identity, repository URLs, and the push destination.

    line 69
  13. Distinctive:

    Update the QA scenarios and advisory coverage.

    line 94
  14. Distinctive:

    Follow the fixture conventions and run the selected scenario with its required prerequisites.

    line 95
  15. Very specific:

    Create or update a source-backed flow doc for every such change.

    line 108

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “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
SKILL.md

Local development

Use this skill for repository development changes. Read the root and applicable nested AGENTS.md, the owning current reference, and relevant source before editing. Keep work within the approved platform scope and preserve unrelated work. Run repository commands from its root; bundled ./ paths below are relative to this skill directory. No global skill installation is required. Use explicit dependency records and focused factories when extracting behavior. The fixture and scenario conventions apply the same composition rules to test infrastructure.

HARD REQUIREMENT: OPEN-SOURCE CONTENT

THIS IS AN OPEN-SOURCE REPOSITORY. NEVER MENTION INTERNAL CODE NAMES OR CORPORATE INTERNAL NAMES IN CODE, COMMENTS, DOCUMENTATION, EXAMPLES, TEST FIXTURES, GENERATED ARTIFACTS, COMMIT MESSAGES, OR PULL REQUEST CONTENT. THIS REPOSITORY SHOULD NEVER CONTAIN CORPORATE INTERNAL NAMES. USE PUBLIC PRODUCT NAMES OR NEUTRAL, DESCRIPTIVE TERMS INSTEAD. CHECK THE ENTIRE PROPOSED CHANGE BEFORE HANDOFF OR PUBLICATION AND REMOVE ANY SUCH REFERENCES. DO NOT COPY INTERNAL CONTEXT INTO THIS REPOSITORY, EVEN AS BACKGROUND OR PROVENANCE.

HARD REQUIREMENT: KEEP IMPLEMENTATION-SPECIFIC BEHAVIOR OUT OF THE CORE

IMPLEMENTATION-SPECIFIC FUNCTIONALITY MUST LIVE IN DRIVERS OR BACKENDS, NEVER BE HARDCODED INTO THE PLATFORM CORE. CORE CODE MUST DEPEND ON PLATFORM CONTRACTS, NOT SPECIAL CASES FOR A PARTICULAR IMPLEMENTATION, VENDOR, OR DEPLOYMENT. EXTEND THE OWNING DRIVER OR BACKEND CONTRACT WHEN NECESSARY; DO NOT BYPASS IT WITH IMPLEMENTATION-SPECIFIC BRANCHES, DEFAULTS, OR DIRECT CALLS IN THE CORE.

IF A DEVELOPER OR AGENT PROPOSES OR INTRODUCES SUCH A VIOLATION, CALL IT OUT EXPLICITLY AND STOP THE SESSION'S IMPLEMENTATION WORK. IDENTIFY THE OFFENDING CODE OR DESIGN, EXPLAIN THE OWNERSHIP BOUNDARY IT VIOLATES, AND ASK FOR A DESIGN CHANGE THAT MOVES THE BEHAVIOR INTO THE APPROPRIATE DRIVER OR BACKEND. DO NOT IMPLEMENT, COMMIT, OR PUBLISH THE VIOLATING APPROACH. RESUME ONLY AFTER THE REVISED DESIGN RESOLVES THE BOUNDARY VIOLATION AND THE USER APPROVES IT.

Workflow

  1. Identify the user-visible outcome, owning primitive, real caller, and affected lifecycle. Inspect the diff and existing docs/flows/ before choosing docs.
  2. Implement the smallest complete change through the owning contract. Preserve the repository's authorization, state, and ownership invariants.
  3. Apply the flow-doc trigger below. For a qualifying change, read the workflow and update the existing behavior flow in the same change. Use the template only when no existing document owns that flow. Use $mermaid-diagrams for diagram semantics. For these flow docs, override its notation defaults: start the Mermaid block with graph TD, omit Mermaid YAML frontmatter, and retain this workflow's required sections so the bundled validator accepts it.
  4. Use $technical-writing when creating, editing, or reviewing documentation. Update affected current references, guides, and navigation. Preserve historical specs and user-owned Manual Notes. Do not create a second source of truth.
  5. Use $enterprise-testing to select proportional checks and satisfy repository integration requirements for new functionality. Follow Add integration tests when adding coverage. For instruction-only changes, check docs, links, skill resources, and any bundled executable; do not run product runtime suites solely for prose. Never run npm run precommit. When changing a default image version, source pin, or bundled runtime version, always complete image version verification before merging. A green ordinary PR check does not replace this workflow.
  6. When publication is authorized, follow the PR policy. Verify identity, repository URLs, and the push destination. Preserve existing remotes; their names do not establish ownership. Preserve the head repository and branch when updating an assigned existing PR. Use the verified base for branch comparisons and reviews. Complete the PR readiness checklist before requesting review and recheck its merge requirements at the final PR head.
  7. Report changed behavior, the flow updated (or a short reason none is needed), checks run, and remaining verification gaps. Do not equate a structural doc check with proof of runtime behavior.

Add integration tests

Register new integration scenarios in the QA matrix:

  1. Extend the supported Agent workflow in tests/integration/qa-matrix-real.test.mjs. Exercise real callers and dependencies, including meaningful failure cases. Use owned disposable fixtures and cleanup.
  2. Register the scenario, applicable presets, and required inputs in tests/helpers/qa-selection.mjs. Update QA manifest for new inputs, files, or preparation; update hosted credential setup when credentials change.
  3. For scenarios in the advisory matrix on main and trusted PRs labeled full-qa, update .github/workflows/qa-advisory.yml. Add the selector to OCC_TEST_QA_SCENARIOS with applicable presets/installations, or add a job for a separate lane. Preserve trusted-PR eligibility and scope credentials to the job's environment.
  4. Update the QA scenarios and advisory coverage. Follow the fixture conventions and run the selected scenario with its required prerequisites. Verify passing cases without skips and complete run-owned cleanup; report any missing proof.

When a flow doc is required

A change is non-trivial when it adds or materially changes a runtime path, state transition, ownership or authorization boundary, persistence behavior, external integration, asynchronous handoff, or consequential decision/failure handling. This includes fixes and refactors that change how these work even if an API signature stays the same. Size and line count do not decide the trigger: a one-line authorization or retry-policy change can qualify.

Create or update a source-backed flow doc for every such change. Cover the changed path and its meaningful boundaries, not every touched file. Prefer a focused update to an existing behavior-named document under docs/flows/; create a new one only for a distinct runtime flow without an existing owner. Link adjacent phases instead of duplicating their traces.

Typos, formatting, copy/link corrections, comment-only edits, behavior-preserving local renames, and test-only or instruction-only maintenance do not require a new flow doc when they leave the documented lifecycle accurate. Correct stale source pointers or flow claims if those edits invalidate them. An explicit request for a flow doc still applies. Do not manufacture runtime documentation for an instruction change merely to satisfy this skill.

Resources and validation

  • Image version verification: mandatory native image checks before merging version changes and publication follow-through.
  • PR readiness checklist: evidence and handoff requirements before review and merge.
  • Flow workflow: source gathering, sections, preservation, provenance, and the required validator command.
  • Flow template: scaffold for new flow docs.
  • Validator: portable structural checks using Python 3.10+ and only its standard library; no package installation needed.

Flow guidance and validator are adapted from Specy 2.0.0. The repository owns this adaptation; it has no dependency on personal skills, memory stores, or session lookup tools. See the developer-skills catalog for provenance.

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.