Add ProseMirror runtime adapter to Evidence Kit

Objective: Add the first non-Slate runtime adapter to the Evidence Kit rich-text benchmark matrix by measuring ProseMirror headless model/state/history operations, wiring the adapter into the active registry, refreshing generated perf artifacts, and proving the adapter gap count decreases with real rows.

Goal plan: docs/plans/2026-05-28-add-prosemirror-runtime-adapter-to-evidence-kit.md

Template: docs/plans/templates/major-task.md

Primary template: docs/plans/templates/major-task.md

Applied packs:

• autogoal
• major-task

Major source:

• type: repo benchmark control plane
• id / link: benchmarks/editor/research/benchmark-registry.json
• title: Evidence Kit rich-text runtime adapter matrix
• decision to make: which non-Slate adapter should land first and what rows it may honestly cover
• decision criteria: source-backed target, headless runtime possible, measured rows produced by an active registry artifact, generated docs and health refreshed

Major lane:

• lane: editor benchmark infrastructure
• output type: adapter implementation plus generated evidence artifacts
• implementation expected: yes
• affected packages / surfaces: benchmarks/editor, generated perf docs, goal plan
• dominant risk: claiming adapter coverage beyond what the headless runtime actually measures

Completion threshold:

• ProseMirror has a required active registry artifact.
• npm run bench:adapter:prosemirror writes at least seven ok rows.
• Rich-text workload coverage marks ProseMirror ok only for the covered headless workload families.
• npm run evidence:refresh refreshes docs, health, and rich-text results.
• Adapter-missing rows drop from 54 to 51; active artifacts rise from 23 to 24; rich-text rows rise from 904 to 911.
• node .agents/rules/autogoal/scripts/check-complete.mjs docs/plans/2026-05-28-add-prosemirror-runtime-adapter-to-evidence-kit.md passes.

Verification surface:

• /Users/zbeyens/git/plate-2/benchmarks/editor
• npm run bench:adapter:prosemirror
• npm run evidence:refresh
• npm run docs:perf:check
• scoped syntax/lint commands recorded below
• generated benchmarks/results/benchmark-health-latest.json
• generated benchmarks/results/rich-text-editors-latest.json

Constraints:

• Start from repo evidence before external claims.
• Keep helper stack proportional.
• Separate measured evidence from source-backed gaps.
• Do not claim browser, React, collab, or product-wrapper parity from a headless ProseMirror adapter.

Boundaries:

• Source of truth: benchmarks/editor/research/benchmark-registry.json.
• Allowed edit scope: Evidence Kit benchmark package, generated perf artifacts, and this goal plan.
• External sources: none used; local ../prosemirror source root exists.
• Browser surface: static generated docs at http://127.0.0.1:8765/index.html.
• Tracker sync: none.
• Non-goals: Lexical, Plate, Tiptap, and ProseMirror DOM adapter work.

Blocked condition:

• Autonomous work would stop only if ProseMirror model/state/history packages were unavailable locally or the adapter could not produce stable measured rows. The adapter ran successfully.

Major state:

• task_type: major
• task_complexity: major
• current_phase: closeout
• current_phase_status: complete
• next_phase: final response
• goal_status: active until final update_goal call

Current verdict:

• verdict: accepted
• confidence: high
• next owner: next adapter task
• reason: ProseMirror now has active measured runtime rows and the matrix still exposes honest remaining gaps.

Completion rule:

• Do not call update_goal(status: complete) until every completion threshold above is satisfied, final evidence is recorded, and the autogoal check-complete command passes.

Start Gates:

Gate	Applies	Evidence
major-task loaded	yes	task classified as heavyweight benchmark adapter work
Active goal checked or created	yes	active goal created for the adapter threshold
Source of truth read before analysis	yes	registry, README, source map, and generated perf output inspected
Major lane selected	yes	editor benchmark infrastructure
Decision criteria stated	yes	source-backed target, active artifact, measured rows, refreshed docs/health
Existing repo patterns / prior decisions checked	yes	prior iterations 002, 003, and 004 pattern followed
Helper stack selected	yes	local Node script plus Evidence Kit registry
External research decision recorded	yes	no web research; local source/dependencies suffice
Implementation expectation recorded	yes	implementation required
Workspace authority selected	yes	/Users/zbeyens/git/plate-2/benchmarks/editor owns commands
Branch / PR expectation decided	yes	no PR requested

Work Checklist:

• Objective includes outcome, completion threshold, verification surface, constraints, boundaries, and blocked condition.
• Major source records source type, id/link, title, decision type, expected outcome, decision criteria, likely files/packages/surfaces, browser surface, and highest-leverage owner.
• Current state is mapped before proposing a new architecture, migration, benchmark, or plan.
• Existing repo patterns, prior decisions, and nearby implementation constraints are recorded before external research.
• External docs or source are used only where repo evidence does not settle the question, or N/A reason is recorded.
• Options, recommendation, tradeoffs, blast radius, and rejection reasons are recorded.
• Facts, inference, and recommendation are separated.
• Review or pressure lenses are selected and completed, or marked N/A with reason.
• Implementation touched the benchmark package and docs/agent evidence surfaces.
• Workspace authority recorded: every proof command names the cwd/tool that owns the analyzed or changed behavior.
• Accepted/actionable review findings are fixed or explicitly rejected with evidence.

Completion Gates:

Gate	Applies	Required action	Evidence
Named verification threshold	yes	Run adapter and refresh commands	npm run bench:adapter:prosemirror; npm run evidence:refresh
Current-state source audit	yes	Map current owner, boundaries, constraints, and affected surfaces	registry, README, source map, iterations inspected
Decision criteria closure	yes	Mark each criterion satisfied, narrowed, rejected, or blocked with evidence	ProseMirror selected for headless runtime coverage only
Options / tradeoffs / rejection record	yes	Record viable options, chosen recommendation, and why alternatives lose	ProseMirror first; Lexical/Plate/Tiptap deferred due larger wrapper/runtime surfaces
Review / pressure pass	yes	Validate no overclaiming against workload coverage	browser/React/collab ProseMirror rows remain adapter-missing
Review findings closure	yes	Fix or explicitly reject accepted/actionable findings and record closure proof	source-map and README updated to stop saying ProseMirror is entirely missing
External-source audit	yes	Cite official/local clone/external sources when used, or record N/A	local ../prosemirror exists; npm ProseMirror packages imported
Implementation gates	yes	Close primary-template and touched-surface gates	adapter, registry, package scripts, generated docs/results updated
Final handoff contract	yes	Record recommendation, evidence, caveats, residual risk, and next owner	see Final handoff contract
Final lint	yes	Run scoped equivalent when files changed	scoped syntax/lint command recorded below
Goal plan complete	yes	Run node .agents/rules/autogoal/scripts/check-complete.mjs docs/plans/2026-05-28-add-prosemirror-runtime-adapter-to-evidence-kit.md	recorded below

Phase / pass table:

Phase	Status	Evidence	Next
Intake and source read	complete	registry/source-map/README/iterations inspected	current-state map
Current-state map	complete	ProseMirror had source target but no measured runtime rows	options
Options and recommendation	complete	ProseMirror chosen as headless runtime adapter; Lexical/Plate/Tiptap deferred	review
Review / pressure pass	complete	coverage limited to three headless workloads	implementation
Implementation or plan artifact	complete	adapter script, registry, package scripts, docs, generated artifacts updated	verification
Verification	complete	focused commands recorded below	closeout
Closeout	complete	final handoff prepared	final response

Findings:

• ProseMirror is the right first non-Slate adapter because its headless runtime maps cleanly to large document creation, text insertion, selection changes, document replacement, fragment insert, and history undo/redo.
• A headless adapter cannot honestly cover DOM/browser/React/collab/product workloads. Those remain visible gaps.
• Evidence Kit can model target-specific runtime coverage through registry runtimeAdapters instead of hard-coding every target in the row generator.

Decisions and tradeoffs:

• Chosen: ProseMirror runtime adapter using prosemirror-model, prosemirror-state, and prosemirror-history.
• Deferred: Lexical, because its meaningful editing path is update/command oriented and needs a separate adapter shape.
• Deferred: Plate and Tiptap, because product-wrapper adapters should measure wrapper overhead instead of pretending to be the same as bare runtime cores.
• Rejected: marking all ProseMirror workload rows ok; that would be fake coverage.

Implementation notes:

• Added benchmarks/editor/benchmarks/prosemirror-runtime-adapter.mjs.
• Added bench:adapter:prosemirror and wired it into bench:evidence, evidence:refresh, and check.
• Added registry runtimeAdapters metadata and required prosemirror-runtime-adapter artifact.
• Updated coverage calculation so measured target adapters can close specific workload rows.
• Refreshed generated perf docs/results.

Review fixes:

• Updated README/source-map wording so ProseMirror is described as partially measured, not entirely missing.
• Kept uncovered ProseMirror browser/product workloads as adapter-missing.

Error attempts:

Error / failed attempt	Count	Next different move	Resolution
None	0	none	no repeated failure

Verification evidence:

• npm install --save-dev prosemirror-model prosemirror-state prosemirror-transform prosemirror-history from /Users/zbeyens/git/plate-2/benchmarks/editor: success.
• npm run bench:adapter:prosemirror from /Users/zbeyens/git/plate-2/benchmarks/editor: success; wrote 7 rows.
• node --check benchmarks/prosemirror-runtime-adapter.mjs && node --check src/index.mjs from /Users/zbeyens/git/plate-2/benchmarks/editor: success.
• npm run evidence:refresh from /Users/zbeyens/git/plate-2/benchmarks/editor: success; generated 911 rich-text rows and health active=24.
• npm run docs:perf:check from /Users/zbeyens/git/plate-2/benchmarks/editor: success; checked index, evidence, rich-text, and internals docs.
• pnpm exec biome check benchmarks/editor/benchmarks/prosemirror-runtime-adapter.mjs benchmarks/editor/src/index.mjs benchmarks/editor/package.json benchmarks/editor/research/benchmark-registry.json benchmarks/editor/README.md benchmarks/editor/research/evidence-source-map.md benchmarks/editor/iterations/005-prosemirror-runtime-adapter.md docs/plans/2026-05-28-add-prosemirror-runtime-adapter-to-evidence-kit.md --fix from /Users/zbeyens/git/plate-2: success; fixed one file.
• Static browser smoke from /Users/zbeyens/git/plate-2: http://127.0.0.1:8765/index.html, rich-text.html, and rich-text-data.json all returned 200.
• Health snapshot: ok=728, adapter-missing=51, coverage-gap=130, optional-missing-artifact=2.
• ProseMirror workload coverage snapshot: ok for core-rich-text-operations-compare, clipboard-large-payload, and history-compare; adapter-missing remains for React, browser, collab, core-current/core-compare, and issue #6038 workloads.

Final handoff contract:

• Recommendation: next runtime adapter should be Lexical, because ProseMirror now covers the clean headless baseline and Lexical is the next distinct runtime architecture.
• Confidence: high.
• Evidence: active ProseMirror artifact, generated rows, refreshed docs/health, and explicit workload coverage map.
• Tests / commands: see Verification evidence.
• Browser proof: static generated docs are refreshed for http://127.0.0.1:8765/index.html.
• PR / tracker: no PR requested.
• Caveats: no DOM/browser ProseMirror adapter yet; no Lexical/Plate/Tiptap adapter yet.
• Next owner: adapter task for Lexical or ProseMirror DOM, depending whether the next question is runtime architecture or browser/product behavior.

Timeline:

• 2026-05-28T17:57:31Z Major-task goal plan created.
• 2026-05-28T18:03:23Z Evidence Kit docs refreshed with ProseMirror adapter rows.
• 2026-05-28T18:05:47Z Plan closeout updated.

Reboot status:

Question	Answer
Where am I?	Closeout
Where am I going?	Final verification, goal completion, response
What is the goal?	Add the next non-Slate runtime adapter
What have I learned?	ProseMirror can close three headless workload gaps without overclaiming browser/product coverage
What have I done?	Added adapter, registry metadata, scripts, generated evidence, docs, and iteration note

Open risks:

• ProseMirror DOM/editor-view behavior is not measured by this adapter.
• Remaining non-Slate adapter gaps are still the dominant next action.