v0.3.0

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog.

[Unreleased]

[0.3.0] - 2026-05-30

Added (0.3.0)

• docs/RUNTIME-PLUGIN-DECISION.md: architect decision record for merging the runtime/plugin track now and deferring broader memory/compression work to follow-up issues
• examples/cross-platform/runtime-file-backed-plugin.mjs: concrete project-local runtime/plugin path with file-backed artifacts and resume support
• scripts/validate-portable-spec.mjs: portable runtime contract validator for project-only, file-backed execution
• scripts/validate-runtime-plugin.mjs: smoke-test suite for isolation, determinism, replay/resume, and path confinement
• Runtime/plugin eval coverage in evals/evals.json and richer benchmark outputs in scripts/grade-evals.mjs and scripts/scaffold-evals.mjs

Changed (0.3.0)

• README cross-platform section: now points to the concrete file-backed runtime/plugin implementation and the merge decision record
• examples/cross-platform/portable-skill-spec.json: version bumped to 0.3.0 and now declares the runtime contract for project-only, file-backed, and forbidden shared-state execution
• examples/cross-platform/ADAPTER-CONFORMANCE-CHECKLIST.md and examples/cross-platform/adapter-conformance-report.template.json: now include project-isolation and shared-state rejection checks
• evals/EVALUATION.md: now states the runtime/plugin merge gate and benchmark deltas for quality, determinism, and replay/resume coverage

Fixed (0.3.0)

• .github/workflows/ci.yml: now validates tooling scripts, the portable runtime contract, and the runtime plugin smoke-test path on master

[0.2.0] - 2026-05-29

Fixed (0.2.0)

• scripts/estimate-cost.mjs: Corrected double-counting of parallel/fan-out agents in assumedCalls.
• scripts/estimate-cost.mjs: Resolved bug where map-based dynamic parallels were incorrectly classified as fixed-count due to parsing precedence.

Added (0.2.0)

• assets/examples/software-dev-pipeline.js: full SDLC workflow with
  spec-intake → recall → plan → execute → verify → persist cycle
• assets/templates/workflow-spec.template.md: topology-first design spec
  template with sign-off gate
• assets/templates/memory-entry.schema.json: stable JSON Schema for scoped
  memory entries
• assets/templates/memory-index.example.jsonl: example memory store content
• scripts/validate-workflow-spec.mjs: linter for workflow design specs
• scripts/estimate-cost.mjs: static pre-run cost estimator for workflow files
• scripts/grade-evals.mjs: automated evaluation grading script
• Patterns 13-16 in references/patterns.md: GSD spec-driven pipeline,
  two-stage review, pressure-test, memory-augmented pre-flight
• Pattern 17 in references/patterns.md: generalized MCP tool call in a leaf
  agent() stage for deterministic, portable MCP integration
• examples/cross-platform/runner-kilocode-runbook.md: Kilo Code adapter
  runbook
• examples/cross-platform/runner-gumloop-runbook.md: Gumloop adapter runbook
• OpenCode adapter coverage in conformance checklist
• examples/use-cases/README.md: six high-value use case scenarios with
  pattern recommendations
• docs/FRAMEWORK-COMPARISON.md: polyflow vs Superpowers vs Vanilla
  Spec-Driven comparison
• docs/DEVELOPMENT-PLAN.md: architecture and implementation plan document

Changed (0.2.0)

• SKILL.md: added topology-first design gate (Step 2 hard checkpoint), scope
  discipline section, state externalization guidance, scoped memory contract
  guidance, Nyquist verification principle
• references/patterns.md: expanded from 12 to 17 patterns
• assets/examples/README.md: updated from six to seven examples
• examples/cross-platform/ADAPTER-CONFORMANCE-CHECKLIST.md: added Kilo Code,
  Gumloop, and OpenCode rows
• scripts/validate-workflow.mjs: added loop safety checks, result filtering
  warnings, and phase title consistency checks

[0.1.0] - 2026-05-23

Added (0.1.0)

• Cross-platform support: portable spec, adapter conformance checklist, and
  per-platform runbooks for Claude Code, Codex, Copilot, and Gemini
• scripts/validate-workflow.mjs linter
• scripts/scaffold-evals.mjs eval scaffolding generator
• Six complete example workflows in assets/examples/
• Starter evaluation test cases in evals/evals.json
• CONTRIBUTING.md, CODE_OF_CONDUCT.md, SECURITY.md
• GitHub issue templates and CI workflow

v0.2.0

2026-05-29

Fixed

• scripts/estimate-cost.mjs: Corrected double-counting of parallel/fan-out agents in assumedCalls.
• scripts/estimate-cost.mjs: Resolved bug where map-based dynamic parallels were incorrectly classified as fixed-count due to parsing precedence.

Added

• assets/examples/software-dev-pipeline.js: full SDLC workflow with
  spec-intake → recall → plan → execute → verify → persist cycle
• assets/templates/workflow-spec.template.md: topology-first design spec
  template with sign-off gate
• assets/templates/memory-entry.schema.json: stable JSON Schema for scoped
  memory entries
• assets/templates/memory-index.example.jsonl: example memory store content
• scripts/validate-workflow-spec.mjs: linter for workflow design specs
• scripts/estimate-cost.mjs: static pre-run cost estimator for workflow files
• scripts/grade-evals.mjs: automated evaluation grading script
• Patterns 13-16 in references/patterns.md: GSD spec-driven pipeline,
  two-stage review, pressure-test, memory-augmented pre-flight
• Pattern 17 in references/patterns.md: generalized MCP tool call in a leaf
  agent() stage for deterministic, portable MCP integration
• examples/cross-platform/runner-kilocode-runbook.md: Kilo Code adapter
  runbook
• examples/cross-platform/runner-gumloop-runbook.md: Gumloop adapter runbook
• OpenCode adapter coverage in conformance checklist
• examples/use-cases/README.md: six high-value use case scenarios with
  pattern recommendations
• docs/FRAMEWORK-COMPARISON.md: polyflow vs Superpowers vs Vanilla
  Spec-Driven comparison
• docs/DEVELOPMENT-PLAN.md: architecture and implementation plan document

Changed

• SKILL.md: added topology-first design gate (Step 2 hard checkpoint), scope
  discipline section, state externalization guidance, scoped memory contract
  guidance, Nyquist verification principle
• references/patterns.md: expanded from 12 to 17 patterns
• assets/examples/README.md: updated from six to seven examples
• examples/cross-platform/ADAPTER-CONFORMANCE-CHECKLIST.md: added Kilo Code,
  Gumloop, and OpenCode rows
• scripts/validate-workflow.mjs: added loop safety checks, result filtering
  warnings, and phase title consistency checks

v0.1.1

Release summary:\n- Add Pattern 17: generalized MCP tool call in a leaf agent stage\n- Add changelog entry for the Wave 2 delta\n\nCommit: a1e534e