Changelog

All notable changes to Spec-Kit are documented in this file.

The format follows Keep a Changelog, and this project uses Semantic Versioning once it reaches 1.0.0.

[Unreleased]

[1.0.0] - 2026-10-31

Added

• Core Pipeline Skills (8 skills with evals):

  • ideator — Divergent idea generation (8-12 options, assumption + cheapest test + FACT/INFERRED/SPECULATIVE labels)
  • buyer-discovery — Outreach tracking → interviews → signal counting + Segment Validation Matrix
  • compliance-health — BPOM/FTC rule engine, risk scoring (0-100), approved/banned claims DB
  • economics-engine — Unit economics calculator with verdict engine, segment analysis, sensitivity analysis
  • plan-compiler — Auto-compiles 11 stages → plan.md with verdict + pivot docs
  • assess — Dimension assessments (business, marketing, financial, product, ops) with grades A+-D
  • pipeline — Master orchestrator: 15 stages + Pivot Decision Gate + resume-from-stage
  • youtube-automation — 7-agent video pipeline: Research → Fact-check → Script → Visuals → Voice → Render → Review/Publish

• Oracle Harness (scripts/oracle_runner.py):

  • Golden test recording/replay for critical skills
  • Function-level call vectors
  • Replay against alternate implementations

• Fail-Soft Tool Chains (skills/fallback_chains.py):

  • Web Search: Exa → Jina → DuckDuckGo
  • Web Fetch: Jina Reader → Exa → Browser
  • LLM: Gemini → Groq → OpenRouter → Pollinations
  • Visuals: Pexels → Pixabay → NASA → Openverse → AI-generated
  • Transcription: Groq Whisper → OpenAI Whisper → Local Whisper
  • Thumbnail: 4 styles as fallback chain

• Quality Gates (.plaesy/instructions/quality-gates.md):

  • 7 blocking business validation gates (B1-B7)
  • 9 pipeline stage gates (S1-S9)
  • Adversarial verifier pattern

• Success Contracts (.plaesy/instructions/success-contract.md):

  • 8 contract templates (Idea Discovery, Buyer Discovery, Economics, Pivot Gate, Plan Compilation, Assessment, Full Pipeline, YouTube Automation)

• Plugin Manifests (multi-platform):

  • .claude-plugin/plugin.json — Claude Code
  • .codex-plugin/plugin.json — OpenAI Codex
  • .agents/plugins/marketplace.json — Agents SDK

• CI/CD (.github/workflows/ci.yml):

  • Evals gate (all 8 skills)
  • Oracle harness replay (4 critical skills)
  • Plugin manifest validation
  • Skill structure validation
  • Fallback chain import test
  • Documentation check

• Documentation:

  • docs/ARCHITECTURE.md — System architecture, data flow, stage mapping
  • docs/DESIGN.md — Design system, stage templates, error states
  • docs/CLI-CONTRACT.md — Complete CLI specifications for all 8 commands

• Using Plaesy Updates:

  • Skill registry with 8 spec-kit skills
  • Pipeline stage map with 15+1 stages
  • Routing table with pivot gate
  • Phase mapping (9-phase ladder + 15-stage pipeline)
  • Quality gates with B1-B7 + S1-S9
  • Success contracts with 8 templates

Changed

• Renamed skills from spec-kit-* prefix to clean names (ideator, buyer-discovery, etc.)
• Updated using-plaesy routing to use clean skill names
• Extended pipeline to 16 stages (added YouTube Automation as optional Stage 16)
• Added Pivot Decision Gate as blocking gate between Stage 9 and 10

Fixed

• Quality gate B3 now properly blocks pipeline until pivot decision documented
• Oracle harness records function-level call vectors for replay
• Fail-soft chains properly skip unavailable tools (no API key = skip)

---

[0.0.1] - Initial release

• Initial release of the Spec-Kit framework: prompts, instructions, agents/roles, checklists, templates, and automation scripts.