agentleFS
peterbamuhigirepeterbamuhigire/chwezi-dev-engine/skills/sdlc-meta/kaizen-improvement-system/SKILL.md

kaizen-improvement-system

Skill31 starsChanged 18 days agoMITView on GitHub

Use when auditing or improving this engineering engine or a product it produces. Coordinates evidence-backed baselines, small experiments, standardisation, and re-audits without replacing domain skills.

2 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

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    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 to turn observed defects, incidents, feedback, source changes, or audit gaps into small, reversible, evidence-backed …

    line 13
  2. Distinctive:

    Teach and re-measure: Update discoverability and handoffs, re-score the changed dimensions, and leave the next improvement visible.

    line 49
  3. Distinctive:

    Do not award 70+ without extraordinary, specific proof or claim production readiness without executable or rendered evidence.

    line 70
  4. Very specific:

    Every Kaizen cycle must begin with digital-research-engine source evaluation and source verification.

    line 116
  5. Very specific:

    Read and search are required.

    line 149
  6. Distinctive:

    If a validator, source, user result, render, system environment, or independent review is unavailable, keep the gap visible and state its …

    line 156

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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
SKILL.md

Kaizen Improvement System

Use this skill to turn observed defects, incidents, feedback, source changes, or audit gaps into small, reversible, evidence-backed improvements. It keeps domain judgement with the owning engine and makes learning discoverable.

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

Use When

  • Auditing or grading the engineering catalogue or a product made with it.
  • Turning defects, incidents, user feedback, or new evidence into a tested improvement.
  • Planning a major iteration, quarterly engine review, or post-release re-audit.

Do Not Use When

  • A single-skill safety review is the only task; use the skill safety gate in skill-engine-audit.
  • A current external claim is needed before source evaluation and verification.
  • The proposed change is cosmetic and has no measurable effect on routing, quality, safety, or delivery.

Required Inputs

  • Engine or product path, intended audience, output types, owner, and constraints.
  • Current routes, references, tests, user or product evidence, and known failure modes.
  • Baseline scoring method, improvement target, acceptance evidence, rollback condition, and re-audit date.

Prerequisites

  • Read docs/continuous-improvement/kaizen-adoption-2026-08.md.
  • For portfolio or current-standards claims, use the Digital Research engine's source-evaluation and source-verification routes.

Workflow

  1. Observe. Inspect the real route, output, users, constraints, evidence, and failure mode.
  2. Baseline. Inventory applicable dimensions and score them with named evidence. Keep raw scores, blockers, uncertainty, and NOT ASSESSED items visible. Publish the repository policy cap of min(raw_score, 65) for permanent audits.
  3. Select. Choose the smallest high-leverage root cause. Prefer repeated defects, user harm, safety issues, missing proof, and routing failures over added prose.
  4. Experiment. Name the exact skill, reference, fixture, validator, or gate; hypothesis; owner; measure; risk; rollback; stop condition; and re-audit date.
  5. Check. Re-run relevant validators, tests, source checks, user or product checks, and independent review. Preserve negative evidence.
  6. Standardise. Promote a successful result into the owning skill, reference, template, fixture, router, or release gate. Do not promote a one-off observation.
  7. Teach and re-measure. Update discoverability and handoffs, re-score the changed dimensions, and leave the next improvement visible.

At retrospective, ask which evidence, repository context, stakeholder access, runtime capability, or tool result was unavailable to the work. Separate process failure from missing access; improve the workflow only where a changed method can alter the outcome.

Core content

For a product, assess the applicable combination of requirements, architecture or document correctness, security and privacy, accessibility, reliability, performance, user value, deployment or handoff, rollback, and evidence quality. For an engine, assess doctrine, taxonomy and routing, skill depth, applied proof, standards currency, output readiness, hygiene, safety, and integrity. For client-facing web or dashboard work, also load references/purpose-fit-premium-ux-implementation.md and require an implementation slice that preserves the client's design thesis rather than copying a reference product.

Quality Standards

  • Scores cite concrete evidence and a one-line deficiency statement.
  • A target of 95/100 is an improvement target, not an achievement claim.
  • Do not award 70+ without extraordinary, specific proof or claim production readiness without executable or rendered evidence.
  • Structural, behavioural, render, system, and production evidence are labelled separately.
  • Missing execution, source, approval, or artefact evidence is NOT ASSESSED.
  • A safety, legal, financial, privacy, security, or release blocker remains a blocker regardless of score.
  • Lessons are stored only when repeatable, generalisable, significant, actionable, and likely to prevent recurrence.

Anti-Patterns

  • Adding a skill without a demonstrated routing or ownership gap. Fix: inventory and test the collision first.
  • Awarding a high score because headings or checkboxes exist. Fix: require outcome-level evidence.
  • Treating a green unit or structural test as product proof. Fix: add the smallest relevant behavioural, failure, render, or handoff check.
  • Closing an action because prose changed. Fix: close only when named acceptance evidence exists and a fresh agent can find the result.
  • Hiding an unavailable check inside an overall score. Fix: mark it NOT ASSESSED with an owner and date.
  • Keeping every retrospective note forever. Fix: promote durable learning and remove redundant narrative.
  • Treating a polished screenshot as a premium implementation. Fix: test real content, critical states, responsive reflow, keyboard/focus, reduced motion, performance, and recovery.
  • Copying a recognisable competitor surface. Fix: implement the user principle and the client's visual thesis, then retain a rationale and originality review.

Outputs

ArtifactConsumed byTemplate
Capped scorecard and evidence registerMaintainer and reviewerKaizen adoption plan and local report
Improvement action recordOwner and implementerGap, root cause, change, hypothesis, measure, risk, rollback, evidence
Standardisation recordFuture agents and routerSkill, reference, fixture, validator, or gate change
Re-audit handoffNext reviewerDate, owner, remaining gaps, and required checks

Evidence Produced

CategoryArtifactFormatExample
CorrectnessBaseline and re-scoreMarkdown tableRaw dimensions, cap, blockers, and evidence
SecuritySafety and integrity reviewMarkdownPermission, provenance, and harmful-output findings
OperabilityRollback and re-audit handoffMarkdownOwner, stop condition, recovery, and next date
Release evidenceValidator and test recordCommand output or MarkdownPositive, negative, and unavailable checks

Read next

  • skill-engine-audit for the engine rubric and evidence matrix.
  • skill-engine-audit safety gate for changed or imported skill safety review.
  • advanced-testing-strategy for risk-scaled validation.
  • world-class-engineering for delivery gates and operational handoff.

Mandatory Digital Research currentness gate

Every Kaizen cycle must begin with digital-research-engine source evaluation and source verification. Record scope, dates, freshness class, support status, uncertainty, and review date for current standards, APIs, frameworks, packages, security, platform, and lifecycle claims; quarantine unsupported claims as NOT_ASSESSED. Apply the currentness gate at digital-research-engine/docs/continuous-improvement/kaizen-currentness-gate.md (resolve that engine per device).

Include claims using superlatives such as “largest”, “first”, “only”, “most severe”, and emergency severity scores in the currentness preflight. Verify severity and scores against the issuing authority; support historical comparisons with a defined comparison set and dated evidence. If either is missing, narrow the claim and record it as unsupported instead of repeating it.

References

  • docs/continuous-improvement/kaizen-adoption-2026-08.md - local adoption plan.
  • digital-research-engine/docs/continuous-improvement/portfolio-kaizen-standard-2026-08.md (resolve that engine per device) - cross-engine standard.
  • skills/sdlc-meta/skill-engine-audit/ - audit dimensions, scoring, and report structure.
  • skills/sdlc-meta/skill-engine-audit/references/skill-safety-gate.md - safety and source-ingestion gate.
  • skills/sdlc-meta/kaizen-improvement-system/references/purpose-fit-premium-ux-implementation.md - implementation patterns for friction reduction, dashboards, state feedback, and authored visual systems.
<!-- dual-compat-end -->

Inputs

ArtefactProduced byRequired?Why
Engine or product path and output typesRouter or task owneryesFixes the audit boundary
Current evidence and failure historyValidators, tests, users, or operatorsyesPrevents opinion-only scoring
Target and constraintsTask owneryesDefines the experiment and rollback

Capability contract

Read and search are required. Execution is preferred for validators and tests; editing requires an authorised improvement scope. Without execution, report the exact checks as NOT ASSESSED. Network research is optional and must use source-disciplined routes.

Degraded mode

If a validator, source, user result, render, system environment, or independent review is unavailable, keep the gap visible and state its consequence. Return a conditional plan or verdict; never convert unavailable evidence into a pass.

Decision rules

ConditionActionFailure avoided
A repeated defect has a clear owning skill or gatePatch that source of truth and add a focused regression checkSymptom-only fixes and recurrence
The proposed change adds only prose or a duplicate routeDo not change the catalogue; record the gap or route to the existing ownerKaizen bloat and routing noise
The result changes a cross-engine handoffRequire source owner, receiving owner, artefacts, uncertainty contract, and fallbackSilent ownership drift
A material outcome, render, system, or production check cannot runMark it NOT ASSESSED and retain the blockerFalse readiness
The experiment passes its acceptance evidenceStandardise the result and schedule re-measurementLearning lost in a one-off patch

The book-wave operating contract is book-driven system decision and agent orchestration. Current technology and security claims must pass Digital Research verification.

The 2026-09-02 cross-engine study is Book-driven Kaizen Wave 3; use it for currentness-gated contracts, testing, Git recovery, secure input handling, protocol semantics, and agent evaluation.

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.