agentleFS
Sign inStart free
peterbamuhigirepeterbamuhigire/proposal-skills/skills/meta/skill-writing/SKILL.md

skill-writing

Skill8 starsChanged 14 days agoMITView on GitHub

Use when creating or normalising reusable proposal skills, trigger routes, contracts, references, or authoring automation under the canonical chwezi-dev-engine skill-writing standard; use skill-safety-audit instead for a read-only security review.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use skill-safety-audit for a read-only inspection of unsafe instructions or bundled resources.

    line 18
  2. Distinctive:

    Read the canonical standard, then this engine's delta.

    line 25
  3. Distinctive:

    Write the input, output, evidence, capability, degraded-mode and decision contracts before the procedure.

    line 26
  4. Very specific:

    Run python -X utf8 scripts/validate_skills.py, python -X utf8 scripts/routing_smoke_test.py, python -X utf8 …

    line 27
  5. Distinctive:

    Stop on any finding or routing collision.

    line 28
  6. Distinctive:

    Every new or changed skill gets positive, negative and collision routing fixtures.

    line 42
  7. Distinctive:

    No book extractions or copied third-party text.

    line 44
  8. Very specific:

    Follow the proposal authoring standard.

    line 47
  9. Distinctive:

    Run the source-ingestion guardrail whenever a book or other copyrighted source informs the work, and apply anti-AI slop before release.

    line 48
  10. Distinctive:

    Read and search are required.

    line 50
  11. Very specific:

    If the canonical standard is unavailable, apply the portable minimum, return the narrowest qualified result, and mark each canonical-only …

    line 52
  12. Very specific:

    When financial-proposal overlaps work-plan, keep pricing and fee assumptions in financial-proposal and route staffing days to work-plan.

    line 64

“Common” is measured: kept by at least 6 in 10 of the 7 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

Skill Writing

Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.

Pointer stub. The canonical standard is chwezi-dev-engine/skills/sdlc-meta/skill-writing (canonical on GitHub; resolve the local checkout through the engine-routing table). Load it first; this file keeps a portable minimum and this engine's delta.

<!-- dual-compat-start -->

Use When

  • Creating or upgrading an active proposal skill, its trigger description or its directly linked resources.

Do Not Use When

  • Use skill-safety-audit for a read-only inspection of unsafe instructions or bundled resources.

Required Inputs

ArtefactSource/providerRequired?If absent
Reusable problem, trigger prompts and neighbour descriptionsRequester and live catalogueYesStop; search the catalogue before drafting.
Canonical skill-writing standardchwezi-dev-engine checkout or GitHubYesApply the portable minimum and mark canonical-only checks NOT ASSESSED.

Workflow

  1. Read the canonical standard, then this engine's delta; inspect the closest neighbours.
  2. Write the input, output, evidence, capability, degraded-mode and decision contracts before the procedure.
  3. Run python -X utf8 scripts/validate_skills.py, python -X utf8 scripts/routing_smoke_test.py, python -X utf8 scripts/source_ingestion_guardrail.py and git diff --check, then python -X utf8 skills/meta/skill-writing/scripts/quick_validate.py <skill-dir>.
  4. Stop on any finding or routing collision; recover by fixing the named contract and rerun, never by weakening the gate.

Outputs

ArtefactConsumerAcceptance condition
Skill directory and routing fixturesMaintainer and routerValidators pass and the expected skill ranks in the top three.

Evidence Produced

EvidenceArtefact and formatConsumerAcceptance condition
Validation and routing recordCommand outputRelease ownerZero findings; unrun checks marked NOT ASSESSED.
<!-- dual-compat-end -->

Quality Standards

  • Portable minimum, applied even when the canonical is unreachable: frontmatter uses only approved keys and name matches the folder.
  • The description starts Use when, stays within 350 characters and names a neighbour, with no workflow steps.
  • SKILL.md stays within 500 lines; deep detail sits in references one level deep, linked directly.
  • Every new or changed skill gets positive, negative and collision routing fixtures.
  • Bundled scripts run through their interpreter, for example python -X utf8 scripts/<name>.py.
  • No book extractions or copied third-party text; paraphrase and attribute.
  • British English, the imperative mood, and NOT ASSESSED for any check not run.

Engine-Local Delta

  • Follow the proposal authoring standard; frontmatter holds only name, description and metadata, and the acknowledgement sits under the title.
  • Run the source-ingestion guardrail whenever a book or other copyrighted source informs the work, and apply anti-AI slop before release.

Capability Contract

Read and search are required. Editing files and running validators need explicit permission for the authoring task; publishing, deletion and release changes need separate authorisation.

Degraded Mode

If the canonical standard is unavailable, apply the portable minimum, return the narrowest qualified result, and mark each canonical-only check NOT ASSESSED; never report it as passed.

Decision Rules

ConditionActionFailure or risk avoided
An existing skill owns the trigger and outputNormalise it in place; put branch-only detail in a linked referenceDuplicate routes and oversized entrypoints

Anti-Patterns

  • Copying the canonical body into this engine. Fix: link the canonical and keep only the delta here.
  • Writing only positive triggers. Fix: name the neighbour and add a collision fixture.
  • Treating an unrun validator as a pass. Fix: record NOT ASSESSED with the reason.
  • Granting edit rights to a review procedure. Fix: default review and audit to read-only.
  • Weakening a baseline to clear a finding. Fix: repair the named contract instead.

Worked Example

When financial-proposal overlaps work-plan, keep pricing and fee assumptions in financial-proposal and route staffing days to work-plan; a fixture asking for a price schedule must rank the financial skill in the top three.

References

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.

Check your setup, free