evidence-control-plane-hard-cut

Objective: Replace the active editor benchmark workflow with an Evidence Kit control plane, registered active benchmark artifacts only, health/next-action reporting, refresh command, generated dashboard cockpit, and docs that say old one-off benchmarks are discarded unless registered.

Completion threshold:

• Active benchmark artifacts are declared in benchmarks/editor/research/benchmark-registry.json.
• benchmarks/editor/src/index.mjs ingests active artifacts from the registry instead of hardcoded Slate v2 artifact lists/fallbacks.
• A health command writes benchmarks/results/benchmark-health-latest.json with status counts, missing/stale/ignored artifacts, and ranked next actions.
• Package scripts expose evidence:refresh and include health generation in full/check flows.
• docs/perf/index.html surfaces benchmark health and next actions.
• cd benchmarks/editor && npm run check passes.
• Served /index.html contains benchmark health content.
• node .agents/rules/autogoal/scripts/check-complete.mjs docs/plans/2026-05-28-evidence-control-plane-hard-cut.md passes.

Verification surface:

• benchmarks/editor/package.json
• benchmarks/editor/src/index.mjs
• benchmarks/editor/benchmarks/benchmark-health.mjs
• benchmarks/editor/research/benchmark-registry.json
• generated benchmarks/editor/docs/perf/*
• cd benchmarks/editor && npm run check
• HTTP smoke proof against http://127.0.0.1:8765/index.html

Constraints:

• Do not delete Slate v2 source-owned benchmark runners; they are the measurement owners.
• Discard old benchmark authority by ignoring unregistered artifacts, not by treating every historical tmp JSON as active.
• Do not revive the old Next/Vite benchmark app lab.
• No PR/commit/branch requested.

Boundaries:

• Source of truth: user instruction "go, discard old benchmarks" plus existing Evidence Kit replacement docs.
• Allowed edit scope: benchmarks/editor, generated perf docs, root benchmark script aliases, and this plan.
• Browser surface: local static docs server at http://127.0.0.1:8765/.
• Tracker sync: N/A.
• Non-goals: no new Slate v2 benchmark runner implementation, no non-Slate runtime adapter implementation.

Blocked condition:

• Block only if the registry-driven ingest cannot preserve current measured rows, required artifacts are missing, package checks fail after targeted repair, or the static docs server is unavailable. None remained.

Task state:

• task_type: benchmark workflow hard cut
• task_complexity: major
• current_phase: closeout
• current_phase_status: complete
• next_phase: final response
• goal_status: ready to close

Current verdict:

• verdict: complete
• confidence: high
• next owner: user
• reason: registry-driven ingest, health report, refresh command, docs cockpit, package check, and served route proof are all green.

Completion rule:

• Close the active goal only after this plan passes the autogoal checker.

Start Gates:

Gate	Applies	Evidence
Skill analysis before edits	yes	Loaded major-task, hard-cut, and autogoal.
Active goal checked or created	yes	Active goal created for Evidence Kit control-plane replacement.
Source of truth read before edits	yes	Read current benchmarks/editor scripts, source map, package scripts, Slate v2 benchmark README, and prior memory note.
Tracker comments and attachments read	no	N/A: no tracker.
Video transcript evidence required	no	N/A: no video.
docs/solutions checked for non-trivial existing-code work	no	N/A: benchmark docs/source map and Slate v2 benchmark README were the direct source.
TDD decision before behavior change or bug fix	yes	No TDD; registry/health script checks and package check are the right proof.
Branch decision for code-changing task	yes	No branch action requested.
Release artifact decision	yes	No changeset; private benchmark lab only.
Browser tool decision for browser surface	yes	Used route-level static server proof.
PR expectation decision	yes	No PR requested.
Tracker sync expectation decision	yes	No tracker sync requested.

Work Checklist:

• Objective includes outcome, completion threshold, verification surface, constraints, boundaries, and blocked condition.
• Task source classified with source type, id/link, title, task type, acceptance criteria, caveats, likely files/routes/packages, browser surface, and root-cause layer.
• Required video or screen-recording evidence is cached/read as normalized <video-transcripts> XML, or marked N/A with reason.
• Nearby repo instructions and implementation patterns read before edits.
• Implementation fixes the right ownership boundary, or the narrower choice is recorded with reason.
• Release artifact requirement recorded: N/A, private benchmark lab only.
• Final handoff shape decided: concise summary with workflow command, files, numbers, and verification.
• Branch handling recorded for code-changing work: N/A, no branch requested.
• Local-env-rot retry policy recorded: N/A, no install corruption signal.
• Workspace authority recorded: verification ran in /Users/zbeyens/git/plate-2/benchmarks/editor and route proof hit the local static docs server.
• High-risk note recorded: command-contract change is limited to benchmark lab/root aliases; package check proves the new flow.
• Review/autoreview target selected: N/A, benchmark lab workflow change with full package check coverage.
• Agent-native review decision recorded: N/A, no agent/tooling surfaces changed.

Completion Gates:

Gate	Applies	Required action	Evidence
Named verification threshold	yes	Run package checks and served route proof.	npm run check passed; served /index.html proof passed.
Bug reproduced before fix	no	N/A: workflow replacement.	N/A.
Targeted behavior verification	yes	Verify registry ingest, health output, docs index.	npm run evidence:refresh passed; health has 23 active artifacts and 10 next actions; served index has health content.
TypeScript or typed config changed	no	N/A.	JS syntax checks passed through npm run check.
Package exports or file layout changed	no	N/A.	No barrels.
Package manifests, lockfile, or install graph changed	yes	Run benchmark package check.	cd benchmarks/editor && npm run check passed.
Agent rules or skills changed	no	N/A.	No agent sync.
Workspace authority proof	yes	Run checks in benchmarks/editor.	npm run evidence:refresh, npm run docs:perf:check, and npm run check ran in benchmarks/editor.
Browser surface changed	yes	Route proof for generated index.	HTTP proof: status 200, health/control-plane/discard text present, 2 primary cards, Evidence Kit not primary.
Browser final proof	yes	Record route proof or caveat.	Route-level static proof recorded; no screenshot needed.
CI-controlled template output changed	no	N/A.	No templates.
Package behavior or public API changed	no	N/A.	No changeset.
Registry-only component work changed	no	N/A.	No changelog.
Docs or content changed	yes	Regenerate and check docs.	npm run docs:perf:check and npm run check passed.
High-risk mini gate	yes	Record realistic failure mode and why chosen boundary is right.	Failure mode: random old tmp JSON becomes active evidence again; registry + ignored-artifact health count prevents it.
Agent-native review for agent/tooling changes	no	N/A.	No agent surfaces.
Local install corruption suspected	no	N/A.	No reinstall.
Autoreview for non-trivial implementation changes	no	N/A: scoped benchmark lab workflow.	Full package check and route proof passed.
PR create or update	no	N/A.	No PR.
PR proof image hosting	no	N/A.	No PR.
Tracker sync-back	no	N/A.	No tracker.
Final handoff contract	yes	Fill final handoff fields.	Final handoff fields completed below.
Final lint	yes	Run scoped Biome/check.	npx biome check ... --fix passed after correcting the root package path.
Goal plan complete	yes	Run autogoal checker.	To run after this update.

Phase / pass table:

Phase	Status	Evidence	Next
Intake and source read	complete	Skills, memory, benchmark docs/scripts read.	implementation
Implementation	complete	Registry, health script, refresh scripts, docs/index cockpit, README/source-map/iteration updated.	verification
Verification	complete	evidence:refresh, docs:perf:check, check, and served route proof passed.	closeout
PR / tracker sync	complete	N/A.	final response
Closeout	complete	Plan ready for autogoal checker.	final response

Findings:

• Source-owned benchmark runners stay in Slate v2; active Evidence Kit ingests only registered stable artifacts.
• The old failure mode was treating every tmp JSON or old benchmark note as current evidence.
• Current health: 23 active artifacts, 904 rows, 716 ok, 54 adapter-missing, 130 coverage-gap, 2 optional-missing-artifact, 2 over-budget, and 33 unregistered historical artifacts ignored.

Decisions and tradeoffs:

• Hard cut means unregistered artifacts are ignored/reported, not promoted.
• Health/next-action output is the self-improving layer; runners remain target-owned.
• The dashboard starts with benchmark health because navigation alone did not explain the workflow.

Implementation notes:

• Added research/benchmark-registry.json with active artifacts and workload mappings.
• Replaced hardcoded Slate v2 artifact lists/fallback artifact paths with registry-driven ingest.
• Added benchmarks/benchmark-health.mjs.
• Added evidence:health, evidence:refresh, and root bench:editor:health / bench:editor:refresh aliases.
• Updated generated index.html to show health, next actions, and ignored old artifacts.
• Updated README/source-map/iteration docs to state the hard-cut registry rule.

Review fixes:

• Corrected a mistaken Biome path for the root package and reran scoped formatting from repo root.

Error attempts:

Error / failed attempt	Count	Next different move	Resolution
Biome path used ../package.json from benchmarks/editor, resolving to missing benchmarks/package.json	1	Reran Biome from repo root with explicit paths	Passed.

Verification evidence:

• npx biome check benchmarks/editor/src/index.mjs benchmarks/editor/benchmarks/benchmark-health.mjs benchmarks/editor/benchmarks/render-perf-index.mjs benchmarks/editor/package.json package.json benchmarks/editor/research/benchmark-registry.json benchmarks/editor/README.md benchmarks/editor/research/evidence-source-map.md benchmarks/editor/iterations/003-evidence-control-plane.md --fix passed from repo root.
• node --check src/index.mjs, node --check benchmarks/benchmark-health.mjs, and node --check benchmarks/render-perf-index.mjs passed in benchmarks/editor.
• npm run bench:rich-text:check && npm run evidence:health passed in benchmarks/editor.
• npm run evidence:refresh passed in benchmarks/editor.
• npm run docs:perf:check && npm run check passed in benchmarks/editor.
• Served proof for http://127.0.0.1:8765/index.html: status 200, health/control-plane/discard text present, 2 primary cards, rich-text and internals primary links present, evidence not primary, next action present.

Final handoff contract:

• PR line: N/A.
• Issue / tracker line: N/A.
• Confidence line: High; full benchmark package check and served route proof passed.
• Flow table:
  • Reproduced: N/A.
  • Verified: evidence:refresh, docs:perf:check, check, and HTTP route proof passed.
• Browser check: route-level proof against local static server passed.
• Outcome: Evidence Kit is now the active control plane; unregistered old benchmark artifacts are discarded from active claims.
• Caveat: No new non-Slate runtime adapters were implemented; health now ranks that as a next action.
• Design:
  • Chosen boundary: Evidence Kit registry + health over target-owned runners.
  • Why not quick patch: hardcoded specs keep old-workflow confusion alive.
  • Why not broader change: benchmark runners themselves still belong in source repos.
• Verified: see verification evidence above.

Final handoff / sync:

• PR: N/A.
• Issue / tracker: N/A.
• Browser proof: route-level HTTP proof passed.
• Caveats: non-Slate runtime adapters remain missing by design and are visible in health.

Timeline:

• 2026-05-28T15:54Z Goal created and source read started.
• 2026-05-28T15:56Z Plan created; implementation started.
• 2026-05-28T16:00Z Registry-driven ingest and health script implemented.
• 2026-05-28T16:03Z Refresh/check/docs/route proof passed.

Reboot status:

Question	Answer
Where am I?	Closeout
Where am I going?	Final response
What is the goal?	Evidence Kit becomes the active benchmark control plane and old unregistered benchmarks are discarded.
What have I learned?	Registry-driven ingest is the missing boundary.
What have I done?	Added registry, health, refresh command, docs cockpit, and proof.

Open risks:

• None for the control-plane hard cut. The next real benchmark work is adapter/runtime coverage, which health now ranks explicitly.