Skip to content

RELEASE-v0.9.0.md

Latest commit

 

History

History
96 lines (65 loc) · 4.89 KB

RELEASE-v0.9.0.md

File metadata and controls

96 lines (65 loc) · 4.89 KB

CodexKit v0.9.0 — Skills That Actually Verify Themselves

82 skills. 5 layers. 21 domain-enriched. Community-ready.

What's new

This release transforms CodexKit from a flat collection of prompts into a layered skill operating system where every skill knows how to verify its own output and evolve from real-world usage.

Domain-specific quality gates for 21 core skills

The 20 most impactful skills now ship with three companion files — written by domain experts, not generic templates:

Asset What it does
verification/checklist.md 4C quality gates (Correctness, Completeness, Context-fit, Consequence) with questions specific to each skill's domain
examples/good-output.md Annotated real-world scenarios showing what a well-executed output looks like
examples/common-mistakes.md Three anti-patterns per skill — the mistakes that waste the most time — with explanations and fixes

Example: The codexkit-incident-postmortem checklist doesn't ask generic quality questions — it asks "Does the timeline include detection-to-resolution with timestamps?" and "Are action items assigned with owners and deadlines, not vague 'we should improve' language?"

Executable Tier 3 assets for technical skills

5 technical skills now ship with runnable scripts and starter templates:

  • test-hardening → coverage-gap-finder.sh — multi-language coverage report generator
  • sql-query-builder → query-validator.sql — checks execution plans, NULL traps, cardinality explosions
  • docs-shipper → readme-template.md + changelog-template.md — Keep a Changelog format
  • automation-designer → automation-spec.md — full automation specification template
  • project-governance-pilot → charter-template.md + raci-template.md — project charter + RACI matrix

Evolution system: skills that learn from usage

  • /evolve-skill workflow: When you discover an edge case, the workflow classifies the change type, applies the right version bump, updates the changelog, and validates the result.
  • gotchas-template.md: Accumulate real-world workarounds per skill — each entry captures the symptom, root cause, and fix so the knowledge doesn't get lost.

Smarter validation

  • validate-pack.mjs now lints v2 sections — warns if a skill is missing Quality Criteria, Verification (4C), Edge Cases, or Changelog.
  • Pack stats now show a per-category skill breakdown so you can see which domains are strongest and which need attention.

Before vs After

v0.8.0 v0.9.0
Skill files SKILL.md + agents/ only 3-tier system with verification, examples, scripts, templates
Output verification Manual — hope for the best 4C checklists with domain questions
Learning from mistakes None gotchas.md + /evolve-skill workflow
Validation depth Frontmatter + structure + v2 section lint + category breakdown
Anti-patterns documented 0 60 anti-patterns across 20 skills
Executable helpers 0 7 scripts/templates for 5 technical skills

Community contribution opportunity

61 skills still need Tier 2 enrichment. This is the single highest-impact way to contribute to CodexKit.

How to contribute a Tier 2 enrichment

  1. Pick a skill you know well from the skill list
  2. Create verification/checklist.md — domain-specific 4C questions
  3. Create examples/good-output.md — annotated scenario
  4. Create examples/common-mistakes.md — 3 anti-patterns with fixes
  5. Reference skills/codexkit-skill-template/ for the expected format
  6. Run node scripts/validate-pack.mjs and open a PR

Skills most in need of enrichment

Domain Skills needing enrichment
Finance & Legal fpa-scenario-modeling, fpa-rolling-forecast, audit-readiness-checker, contract-drafter, legal-memo-writer, legal-due-diligence
HR & Operations job-description-writer, performance-review-writer, interview-guide-builder, supplier-evaluation, procurement-rfp-writer
Strategy & Marketing strategic-analysis-porter, competitor-intelligence, campaign-brief-writer
Engineering architecture-decision-writer, api-design-reviewer
Cross-functional decision-log, crisis-communication, policy-document-writer

Deep Research shortcut

Use templates/deep-research-prompt.md with ChatGPT Deep Research, Gemini, or Perplexity Pro to generate the domain knowledge for any skill enrichment. The 6-section output maps directly into the SKILL.md and companion files.

Install / Update

# New install
bash ./START-HERE.sh

# Update existing
bash ./UPDATE.sh

# Validate
node scripts/validate-pack.mjs

Windows users: double-click START-HERE-WINDOWS.cmd or UPDATE-WINDOWS.cmd.

Full changelog

See CHANGELOG.md for the complete list of changes.