Spec Driven Development Feature Matrix

Table of Contents

This matrix compares the seventeen spec-driven development tools profiled in this section, feature by feature.

The category splits on two axes: who owns the specs (your repo or a platform) and whether the ceremony sizes itself to the change, and the waterfall critique is the standing judge of the second axis, while the GSD lineage adds a third question, whether the steward survives their own controversy, and the October additions extend the field to a vendor-backed lifecycle engine, an installable skill set, a company-born governance layer, and a requirements-centric RUP revival, with the prompt-pack column proving the floor is four markdown files and nothing installed, Superpowers closing the set as the mandatory-workflow skill suite holding the category’s largest single adoption figure, and Spec-Driven Develop completing it as the GitHub-native wing, where the tracker is Issues, Milestones, and one batch PR per phase, with the October 21 addition extending the field again to Archcore’s typed-document context layer, whose review names the side that is wrong.

Legend: ✓ supported, ✗ not supported, ~ partial or conditional, ? not verified. Each column links to the full research note; every cell below traces to a source cited there or in the references.

The matrix
#

Feature	AI Unified Process	AI-DLC	Arc Kit	Archcore	BMad Method	cc-sdd	GitHub Spec Kit	GSD	Intent-Router	Mycelium	OpenSpec	PDD Kit	Spec Kitty	Spec-Driven Develop	Superpowers	Tessl	Vibe Coding Prompt Template
Kind	methodology plus agent plugins and skills, aiup-core plus four stack plugins, marketplace or Tessl install	vendor methodology plus native CLI engine, installer per harness	lifecycle command catalog, 76 governance commands as a Claude Code plugin plus six agent formats	Go CLI plus per-agent plugins and MCP access, project context as typed Markdown files under .archcore/	method plus agent workflows, npm install	17-skill SDD workflow installed into each agent’s native skills, npm installer	Python CLI plus slash-command templates	npm command-and-prompt framework, multi-runtime, ecosystem growing around the loop	single installable skill: an intent compiler emitting a typed spec artifact with a delivery check	Claude Code plugin of structural hooks plus 63 optional skills, ports to OpenCode, Codex, and Cursor	spec toolkit CLI, npm install	three agent skills (map, ideate, verify) plus a CLI for lint, local report, and blueprints, no runtime	Python CLI, repo-native mission state with worktree lanes and a governance trail	three pure-Markdown skills plus optional Claude Code, Codex, and OpenCode plugins, no runtime	skill-suite methodology installed per harness, mandatory workflows from session start	hosted platform, thin CLI	copy-paste prompt pack, optional npm CLI
Steward	solo (Simon Martinelli, Java Champion, author of the method’s Apress book)	AWS Labs (methodology by Raja SP, AWS principal solutions architect)	solo maintainer, 95 of the last 100 commits	Archcore (the archcore-ai org, company site and docs, 8 public repositories)	BMad Code, LLC	solo (Gota, gotalab)	GitHub	original archived at TÂCHES (gsd-build), active as Open GSD community	solo (angel291592)	solo (Håvard Bartnes)	Fission AI	solo (Itai Gendler)	Spec Kitty, Inc. (creator Robert Douglass)	solo (zhu1090093659)	Jesse Vincent and Prime Radiant (commercial support)	Tessl (Podjarny), $125M raised	solo (KhazP)
License	~ Apache-2.0 with NOTICE and trademark terms on derived works	✓ MIT-0	~ MIT except one proprietary Claude-only UK G-Cloud overlay	~ Apache-2.0, plain grant, the category’s second non-MIT open license beside AIUP’s	✓ MIT	✓ MIT	✓ MIT	✓ MIT	✓ MIT	✓ MIT	✓ MIT	✓ MIT	✓ MIT	✓ MIT	✓ MIT	✗ closed platform	✓ MIT
Artifact model	vision, requirements catalog (FR/NFR/C), entity model, use-case diagram and per-use-case specs with BR rules, BPMN processes, and journey test cases under docs/, implementation markers in code and test annotations	plans, units, and decisions in repo state, 113-event audit trail, learned team rules	governance documents (principles, business cases, requirements, ERDs, Wardley maps, HLD/DLD reviews) with provenance and citation markers	typed documents under .archcore/ (spec, plan, adr, rule, doc types carried in the filename), linked to each other, versioned with the code	briefs, specs, architecture carried forward	brief, roadmap, requirements/design/tasks per spec with `_Boundary:_` and `_Depends:_` annotations, per-task implementation notes	constitution, spec, plan, tasks files	milestone state persisted as STATE.md, CONTEXT.md, PLAN.md; heavy work in fresh-context subagents	one typed IntentSpec per intent at `.intent/<intent>.intent.yaml`, evidence pointers per probed field, state recorded (waiting, settled, carried out), delivery-check result written back	four-question discovery brief as `.claude/canvas/` YAML (purpose, jobs-to-be-done) with evidence fields (validated, strength), a decision log, and delivery diamonds, plain files in your repo	delta proposals archived into a living ledger	a standing present-tense product model under product-model/ (entities, interfaces, capabilities with scenarios, journeys, business rules, variations) plus coverage.md declaring which code the model accounts for	specs, plans, work packages, review state, and merge decisions under `kitty-specs/`, retrospective per mission	phase analysis with S.U.P.E.R health scores, a batched task plan, MASTER.md progress index, one GitHub Issue per task in GitHub modes, archived governance snapshots	brainstorm design doc, bite-sized task plan with exact paths and verification steps, per-task subagent reviews in isolated git worktrees	specs live on the platform	research findings, PRD, and tech design saved as docs files, then AGENTS.md, MEMORY.md, a review checklist, and agent_docs/ briefs generated by Part 4
Workflow entry	/requirements from docs/vision.md, then /entity-model, /use-case-diagram, /use-case-spec, /spec-review, with a stack plugin owning construction; /reverse-engineer for brownfield starts	`/aidlc <request>`, routed into one of 11 workflow profiles	slash commands per lifecycle phase (/arckit-requirements and 75 more)	/archcore:init proposes or imports the context, then /archcore:plan, /archcore:document, and /archcore:review; plain sentences work in every connected agent	`bmad-build` with right-sized depth	`/kiro-discovery` routes to spec, multi-spec batch, or direct implementation	`/speckit-constitution` then specify	`/gsd-new-project` or `/gsd-onboard`, then discuss-plan-execute-verify-ship per milestone	the request itself: the skill probes, asks, then emits; installed with npx skills add, prompt-only at its L0 backend	`/mycelium:start` asks the four questions once per repo, then the gates fire on their own, with 63 optional skills above them	`/opsx:propose` then apply and archive	`/pdd-map` on an existing repo or `/pdd-ideate` from an idea, then implement with any workflow, then `/pdd-verify` against the model	`spec-kitty dispatch`, then the spec, plan, tasks, next, review, accept, merge pipeline	automatic on large-change phrasing (rewrite, migrate, overhaul) or `/spec-dev <task>`; Deep Discuss for the ambiguous middle, `review-spd` for findings-first review	automatic: the session-start bootstrap triggers brainstorming before any code	web workflow	four prompts pasted into any chat tool, then a build prompt in the coding assistant
Ceremony sizing	~ fixed artifact chain: skills run per artifact but nothing shrinks the ceremony for small changes, and /use-case-spec asks up to five questions per use case	~ Classic, Express, and focused profiles; all 33 stages remain available	heavyweight, strategy through assurance in one catalog, overlays add jurisdiction depth	✓ the package fits the change: a small fix gets no documents, a settled choice a decision record, a large initiative a PRD and one spec per capability	✓ right-sizes to the change	✓ discovery can route to no-spec direct implementation	✗ fixed ceremony	~ same five-step loop each milestone, lighter quick-task mode	✓ the gate sizes itself: a clear request emits in one pass, an ambiguous one stops at a single question	~ the gate sizes itself by repo state: an empty repo stops at the first new source file, an established repo is asked once, ever, and the six-layer depth is chosen at start, but the four questions run only once	~ fixed but light	~ no ceremony of its own: the model scales with the product rather than the change, a purely technical change never touches it, and the docs advise small models for small products	✗ fixed pipeline, and its own docs call it overkill for one-off edits	~ keyword-gated: the seven-phase pipeline triggers on large-scale phrasing and small work never enters it, but every phase runs once it does	~ 7.0’s rebuilt brainstorming sizes the process to the work, from a quick task to a full design, and you can override its call; the skill suite itself still installs as mandatory	? not verified	~ fixed five-step sequence, research optional at the start
Brownfield support	✓ /reverse-engineer recovers entity-model and use-case baselines from an existing codebase	~ bug-fix, infrastructure, and security profiles target existing systems; no explicit brownfield onboarding path	~ traceability and citation commands attach to existing documents and estates	✓ repo-native by design: /archcore:init reads the repository and imports existing CLAUDE.md, AGENTS.md, and rule files	✓ establish-context path	✓ `/kiro-steering` captures existing project context; Kiro specs import	~ not the primary case	✓ `/gsd-onboard` for existing repos	✓ repo-native: the probe reads your deps, routes, git history, ADRs, and tickets before asking	~ `/adopt` populates what an existing codebase can establish and patches the holes, but the four-question brief aims at new product work	✓ explicit design goal	✓ /pdd-map reads an existing repository without running it and writes the model from it, with coverage.md recording what stayed unmapped	~ repo-native so existing code is the default case, no dedicated onboarding path documented	✓ built for rewrites and migrations of existing systems: Phase 1 inventories modules and grades architecture health before planning	~ repo-native (worktrees branch from your project), no dedicated existing-repo onboarding path documented	? not verified	~ starts from an idea in a fresh folder, no existing-repo path
Convergence checking	✓ deterministic spec_lint failing CI, advisory /spec-review findings, and /coverage-check mapping every use-case unit onto code and tests with a justified status (the audit agent ships in two of four stack plugins)	~ human approval gates plus source-bound review evidence at every stage	? not recorded	✓ /archcore:review compares the branch with the documents in both directions and names the side that is wrong: code-wrong or spec-wrong	✓ verify and learn loop	✓ per-task independent reviewer plus `/kiro-validate-impl` GO/NO-GO verdict	✓ converge step	✓ verify phase walks the build before done	✓ built work is checked against the spec one constraint at a time, with the result written back to the artifact	~ nothing checks built code against the brief; the decision log and diamonds record what was decided and the corrections file forces re-reads, the check is human	~ archive keeps ledger current	~ /pdd-verify checks the code against the model and reports every gap, the human chooses what to fix, and lint checks the model’s structure, not the code	✓ review, accept, and merge gates with attached evidence	~ per-task S.U.P.E.R checklist plus tiered review (machine validation, orchestrator diff, independent reviewer for lanes and high-risk work), and a drift score that halts or re-decomposes at thresholds	✓ TDD enforced (code written before its test gets deleted), two-stage per-task review, critical findings block progress	? not verified	~ the build prompt demands code checked against the PRD and the saved checklist drives implementation checks, no verifier beyond the model
Unattended execution	~ implementation skills ask before implementing an unapproved use case; humans hold the Approved decision and the business review each iteration	✗ approval gates require human decisions by design	✗ artifacts are drafts for qualified people to review by its own disclaimer	~ the agent implements from the spec, examples, and tasks, but review is the gate before merge and nothing runs unattended end to end	✓ BMad Loop module	~ autonomous `/kiro-impl` per task, humans hold the phase gates	✗	~ parallel executor waves, human verify step, gsd-loop works a GitHub queue	~ not an executor: the skill hands the spec to your planner or agent, and later sessions read the artifact back	~ the agent builds once a delivery diamond is open; it cannot write its own skip acks or turn off a stop without asking, and a key it writes cannot be committed or pushed	✗	~ not an executor: implementation stays with your agent and its workflow, with an opt-in verify-orchestrated loop of phased implement-and-check	~ agents implement in parallel worktrees, humans hold review, accept, and merge	~ orchestrator-centric tiers run parallel lanes (up to 4) in worktrees with batch PRs, humans hold the Phase 5 confirmation and review gates	~ subagent runs work autonomously for hours at a stretch, humans hold design sign-off and the go decision	? not verified	~ the optional npm CLI pitches agent-driven runs, the prompt flow itself is human-run
Adoption	146 stars, 26 forks (as of 2026-10-17), named reference teams, one independent HN endorsement, one public field report	about 5.1k stars (as of 2026-10-14)	2,255 stars as of 2026-10-14, no HN footprint	71 stars, 13 forks (as of 2026-10-21), no independent coverage found	about 54k stars (as of 2026-10-14)	about 3.7k stars, 25k npm downloads a month (as of 2026-10-21)	about 141k stars (as of 2026-10-14)	original archived at 64.3k stars; successor 10.5k stars (as of 2026-10-14), 43k npm downloads a month (as of 2026-10-21)	239 stars on 2026-10-14, re-inflating after collapsing from 1,017 on 2026-10-11, created 2026-09-22	47 stars as of 2026-10-16, three of the steward’s own projects, a three-person test cohort, and one stranger on record, no independent coverage	about 71k stars (as of 2026-10-14), 2.8M npm downloads a month (as of 2026-10-21)	80 stars, 9 forks (as of 2026-10-23), an npm package ten days old at 32 downloads a month, one 1-point HN submission, no independent coverage	about 1.7k stars, 3.3k PyPI downloads a month (as of 2026-10-21)	986 stars, no release since 2026-07-26 (as of 2026-10-14)	about 299k stars, approaching 300k per its steward (as of 2026-10-14)	24-point raise thread, thin OSS surface	about 3.1k stars (as of 2026-10-14)
Pricing	free Apache-2.0 plugins; workshops CHF 4,400 per in-house day, online from $290; Studio private beta	free	free MIT; one public-but-proprietary supplier overlay	free Apache-2.0; no paid tier found (as of 2026-10-21)	free	free	free	free	free	free	free	free MIT; no paid tier found (as of 2026-10-23)	free CLI; demo-booked platform, no published tiers	free	free; Prime Radiant sells commercial support, no published tiers	free tier plus Team at $100 per month, Enterprise custom	free

Reading the matrix
#

The license and steward rows tell the ownership story: thirteen MIT-licensed repo-native toolkits, two Apache-2.0 open projects (one a methodology with a training business behind it), and one closed, funded platform, and the free tools set the price anchor at zero while Tessl spends $125M betting specs are rentable. The adoption row inverts the funding row, which is the tension to watch. Superpowers moves the adoption ceiling: its 299k stars are more than double Spec Kit’s, and it is the first column whose steward company sells support around a mandatory method rather than platform access. GSD adds the stewardship question the category had not faced: the archived original out-stars every column except Superpowers, Spec Kit, and OpenSpec, and its 10.5k-star successor is rebuilding trust in public. The three newest columns extend the steward axis: AI-DLC is the first column owned by a cloud vendor’s methodology org, cc-sdd is the first owned by one person, and Spec Kitty is the first that began as a derivative of another member’s workflow and now runs as a company product around it. The prompt pack is the first column with no machinery at all, four markdown files and a solo maintainer. Intent-Router moves the floor below the prompt pack: one skill with no files to keep in sync, whose contract is the typed artifact it writes and whose convergence check is the delivery comparison, the only column whose subject is the request itself rather than the delivery process around it. Spec-Driven Develop moves the tracker itself into GitHub: Issues are the acceptance records, one batch PR per phase is the delivery unit, and the drift score is the only execution control in the category that measures how far reality has left the plan and halts at thresholds, though it is also the only column that has gone quiet (no release since July 2026). Mycelium moves the gate earlier than any column: its spec is not a plan for the work but an evidence-graded record of whether the work should happen at all, it is the only column whose hooks hold the agent’s first source file, and it arrives with the category’s emptiest independent footprint, 47 stars and one self-published Show HN thread.

PDD Kit moves the artifact question from the change to the product: where OpenSpec archives change deltas into a ledger of what changed, PDD Kit keeps one present-tense model of what the product is, declares which code it accounts for, and leaves the per-change spec to whatever workflow you already run, making it the first column designed to compose with the others rather than replace them.

Archcore moves the review axis: its verdict names the side that is wrong, code-wrong or spec-wrong, where every other convergence mechanism reports findings for a human to adjudicate, and it joins AIUP as the category’s second Apache-2.0 grant.

AI Unified Process moves the doctrine axis: it is the first column that declares the code the current implementation of the spec rather than the other way around, the first taught from its own Apress textbook, and also the smallest adopted bet in the table at 146 stars concentrated in the German-speaking Java consulting orbit its four stack plugins encode.

Ceremony sizing is the design axis the waterfall critique created: BMad and cc-sdd let small changes skip the ceremony (BMad by design, cc-sdd by discovery routing), Superpowers 7.0 sizes it through brainstorming with an override, Spec-Driven Develop’s keyword trigger keeps small work out of the pipeline entirely, and the artifact-first tools pay for their simplicity with fixed ceremony (GSD at least ships a lighter quick-task mode, while Spec Kitty’s docs just call the pipeline overkill for small work). If your changes are mostly small, that row alone picks your column.

Brownfield is the sleeper row: OpenSpec is explicitly built for existing code, BMad has an establish-context path, GSD’s onboard command covers it, Spec Kitty is repo-native by default without an onboarding path, and spec-kit’s scaffolding still assumes a fresher repo than most of us have, while the prompt pack is the only column that is greenfield by design.

Choosing from the matrix
#

Want requirements, not code, to be the artifact that outlives the system, with audits tying every business rule to a test: AI Unified Process, accepting the four-stack ceiling and the RUP ceremony.
Existing codebase, want the spec ledger to stay current: OpenSpec.
Want a standing description of the whole product that survives every change, with the per-change spec left to the tool you already run: PDD Kit, accepting a young, zero-coverage steward.
Want decisions, specs, and rules kept as typed files a review gate enforces across agents, no whole method attached: Archcore, accepting a 71-star pre-1.0 steward.
Want a whole delivery process with roles and retrospectives: BMad Method.
Heterogeneous org, zero cost, constitution ceremony acceptable: Spec Kit.
Want a whole-lifecycle, vendor-backed engine with gates and an audit trail: AI-DLC, accepting the AWS gravity.
Want review machinery installed as skills in the agent you already run: cc-sdd, on its two stable hosts.
Want a phase loop that quarantines heavy work in fresh-context subagents: GSD, accepting the young successor’s governance.
Want the request itself compiled into a typed contract, with the delivery checked against it, no workflow adopted: Intent-Router, reading its three-run self-eval as a demo, not proof.
Want the agent structurally unable to start a new product before its reasons are written down, with the evidence graded: Mycelium, accepting a 47-star pre-1.0 tool with no independent coverage.
Want parallel agent work in isolated worktrees behind review gates and an audit record: Spec Kitty, accepting the company gravity.
Tracking a large rewrite in GitHub Issues and shipping one reviewable batch PR per phase, with a drift score to catch plan decay: Spec-Driven Develop, accepting a solo steward two months quiet.
Want the process enforced rather than suggested, and accept the measured token-and-time premium: Superpowers.
Want specs as a managed product and accept portability questions: Tessl.
Starting from a blank page with nothing installed: Vibe Coding Prompt Template, accepting that nothing enforces the steps.

Changes
#

2026-08-27 - Created as a single-column Spec Kit scaffold.
2026-08-27 - Rebuilt to four columns (BMad, OpenSpec, Spec Kit, Tessl) across the ownership and ceremony-sizing axes.
2026-09-16 - Refreshed the Spec Kit adoption cell to about 137k stars after the v1.0.7 release.
2026-09-16 - Extended from four to five columns with GSD (the archived get-shit-done lineage plus the active gsd-core successor), inserted alphabetically, every row gaining a cell, and the reading prose extended to the stewardship question.
2026-09-21 - Refreshed the GSD and OpenSpec adoption cells after their star and npm-download counts moved.
2026-09-24 - Removed the verification preamble line on owner request.
2026-09-25 - Refreshed the Spec Kit (about 139k stars) and GSD (successor 9.8k stars, 40.9k npm downloads) adoption cells.
2026-09-27 - Refreshed the GSD (successor 9.9k stars, 41.2k npm downloads) and OpenSpec (2M npm downloads) adoption cells.
2026-09-29 - Refreshed the BMad (about 54k stars), GSD (41.9k npm downloads), and OpenSpec (about 71k stars) adoption cells; all other cells re-verified unchanged.
2026-10-02 - Refreshed the BMad (about 53.7k stars) and GSD (successor 10.1k stars, 43k npm downloads a month) adoption cells; all other cells re-verified unchanged.
2026-10-06 - Extended from five to seven columns with AI-DLC and cc-sdd (inserted alphabetically), every row gaining cells, the reading and choosing prose extended to the vendor-backed and skill-set columns, and the GSD (successor 10.2k stars) and OpenSpec (2.5M npm downloads) adoption cells refreshed.
2026-10-06 - Extended from seven to eight columns with Spec Kitty (inserted alphabetically), every row gaining cells, the reading and choosing prose extended to the governance-and-worktrees column, and as-of qualifiers added to the adoption row.
2026-10-07 - Extended from eight to nine columns with Vibe Coding Prompt Template (inserted alphabetically), every row gaining cells, and the reading and choosing prose extended to the zero-install prompt-pack column.
2026-10-08 - Refreshed the adoption-row as-of dates to 2026-10-08 and the moved cells (AI-DLC about 5.1k, BMad about 53.9k, GSD successor 10.3k stars); all other cells re-verified unchanged.
2026-10-09 - Refreshed the adoption-row as-of dates to 2026-10-09 and the moved cells (cc-sdd about 27k npm downloads, GSD original 64.3k and successor 46k npm downloads, OpenSpec 2.7M npm downloads); all other cells re-verified unchanged.
2026-10-09 - Extended from nine to ten columns with Superpowers (inserted alphabetically), every row gaining cells, and the reading and choosing prose extended to the mandatory-workflow skill suite.
2026-10-10 - Moved the Superpowers ceremony-sizing cell from fixed to partial on the v7.0.0 rebuilt brainstorming, refreshed its adoption cell (about 297k stars), and re-verified the adoption row with as-of dates moved to 2026-10-10 (BMad about 54k stars, cc-sdd 29k npm downloads, GSD successor 10.4k stars and 48k npm downloads, OpenSpec 2.9M npm downloads); all other cells re-verified unchanged.
2026-10-10 - Extended from ten to eleven columns with Intent-Router (the single-skill intent compiler with a typed artifact and a delivery check), sorted between GSD and OpenSpec, every row gaining a cell, and the reading and choosing prose extended to the intent-compiler column.
2026-10-11 - Extended from eleven to twelve columns with Spec-Driven Develop (the GitHub-native tracker: one Issue per task, one batch PR per phase, and a drift score that halts execution), sorted between Spec Kitty and Superpowers, every row gaining a cell, and the reading and choosing prose extended to the GitHub-native column.
2026-10-11 - Refreshed the adoption-row as-of dates to 2026-10-11 and the moved cells (cc-sdd 29k npm downloads, GSD successor 47k npm downloads, Intent-Router 1,017 stars); all other cells re-verified unchanged.
2026-10-13 - Moved the Intent-Router adoption cell after the GitHub API read 1,017 stars and 124 forks on 2026-10-11 and 158 and 46 on 2026-10-13, a drop fork counts cannot take organically, recorded as an apparent engagement purge in the cell itself; all other cells re-verified unchanged.
2026-10-13 - Extended from twelve to thirteen columns with Arc Kit (the 76-command enterprise-architecture governance harness), inserted in sorted position after AI-DLC, all eleven cells traced to the new note, intro count bumped.
2026-10-14 - Corrected the reading prose: the archived GSD original (64.3k stars) does not out-star OpenSpec (about 71k), so OpenSpec joins the columns above it.
2026-10-14 - Refreshed the adoption-row as-of dates to 2026-10-14 and the moved cells (Spec Kit about 141k stars, Superpowers about 299k, cc-sdd 27k npm downloads a month, Arc Kit 2,255, GSD successor 10.5k, Spec-Driven Develop 986), rewrote the Intent-Router cell for its re-inflation (158 to 239 stars), and re-verified the Spec Kitty PyPI figure byte-identical; all other cells re-verified unchanged.
2026-10-16 - Extended from thirteen to fourteen columns with Mycelium (the discovery gate: harness hooks hold the agent’s first source file until a four-question, evidence-graded brief exists), inserted in sorted position between Intent-Router and OpenSpec, every row gaining a cell, the reading and choosing prose extended, and the intro count bumped to fourteen.
2026-10-16 - Corrected the license-prose count while inserting the column (nine to twelve MIT-licensed repo-native toolkits; the count had not moved with the October additions).
2026-10-17 - Extended from fourteen to fifteen columns with AI Unified Process (the requirements-centric RUP revival: use cases as the living source of truth, deterministic spec lint that fails CI, and a coverage audit tying every rule to code and tests), inserted in sorted position before AI-DLC, every row gaining a cell, the intro count bumped to fifteen, and the license prose rewritten to twelve MIT toolkits plus one Apache-2.0 methodology against one closed platform.
2026-10-18 - Refreshed the cc-sdd adoption cell (26k npm downloads a month, as of 2026-10-18) after the rotation re-verification; all other cells re-verified unchanged.
2026-10-19 - Refreshed the GSD (43k npm downloads a month) and OpenSpec (2.8M npm downloads a month) adoption cells; all other cells re-verified unchanged.
2026-10-21 - Extended from fifteen to sixteen columns with Archcore (the Apache-2.0 typed-document context layer whose review names the side that is wrong, code-wrong or spec-wrong), inserted in sorted position between Arc Kit and BMad Method, every row gaining a cell, the AIUP license cell’s only-non-MIT claim dropped, the intro count bumped to sixteen.
2026-10-21 - Refreshed the adoption-row as-of dates to 2026-10-21 and the moved cells (cc-sdd 25k npm downloads a month, GSD 43k npm downloads a month, OpenSpec 2.8M npm downloads a month, Spec Kitty 3.3k PyPI downloads a month re-verified byte-identical); all other cells re-verified unchanged.
2026-10-23 - Extended from sixteen to seventeen columns with PDD Kit (the standing present-tense product model under product-model/, verified against the code by /pdd-verify and composed with any workflow), inserted in sorted position between OpenSpec and Spec Kitty, every row gaining a cell, the intro count bumped to seventeen, the license prose count moved to thirteen MIT toolkits, and the reading and choosing prose extended to the product-model column.