=====================================================================
 G-WARP AI TEAM  -  WHO IS WHO + WORK LOG
 EVERY AI MUST READ THIS FILE FIRST, BEFORE STARTING ANY WORK.
=====================================================================

AI CHAT-CONTEXT PRESERVATION RULE (2026-10-17) — REQUIRED FOR EVERY AI
- Every active AI must keep a chronological conversation/context log inside its OWN folder named <ai-folder>/AI-CHAT-LOG.txt (for example, chatgpt/AI-CHAT-LOG.txt, gemini/AI-CHAT-LOG.txt, grok/AI-CHAT-LOG.txt). The former deepseek/AI-CHAT-LOG.txt path is historical only; DeepSeek is no longer an active role.
- For each task/session, append the user's request accurately and the AI's response/handoff. Include date, task, files changed, evidence/test status, unresolved questions, and exact next step.
- Preserve decisions, constraints, corrections, and explicit prohibitions. Keep direct user wording separate from AI summaries. Never invent missing chat history or claim an incomplete transcript is complete.
- If the platform may truncate/delete history, update the log before ending the session. Read your own log before continuing an existing task.
- Never log passwords, API keys, tokens, private keys, personal secrets, or unnecessary sensitive data. Redact secrets. Keep logs focused on project work.
- If original chat is unavailable, write ORIGINAL CHAT NOT AVAILABLE and only summarize supported facts; ask the owner rather than guessing.
- Use team/AI-CHAT-LOG-TEMPLATE.txt. This rule does not grant cross-folder editing, live integration, testing authority, or release permission.

THE TEAM
- Owner: Creapter1 (the boss - decides priorities, design picks, merges and releases).
- ChatGPT  | Independent reviewer / planner / test designer / evidence reconciler | folder: chatgpt/ | list: team/TODO-CHATGPT.txt | queue: team/CONTINUOUS-WORK-QUEUE.txt
- Grok     | Scratch behavior specifications, expected-result tests and extension audit | folder: grok/ | list: team/TODO-GROK.txt
- Gemini 2 (FreeBuff/Argon access path) | Combined role: original G-Warp visuals + browser extension + lab research, A/B/C candidates, test harnesses and reports formerly assigned to DeepSeek | one active folder: gemini/ | list: team/TODO-GEMINI.txt
- DeepSeek | RETIRED from active work on 2026-10-18. Its old outputs are preserved as legacy material under gemini/lab/; old work-log entries are historical, not current assignments.
- Claude Opus 5.5 | Integration lead + official runtime/visual evidence owner. Reviews all AI work, assigns bounded revisions/assets, integrates approved candidates, captures screenshots, and prepares releases. Must verify active model identity and follow owner approval gates.

START HERE: team/WORKFLOW-START-HERE.txt, then TODO.txt and team/CONTINUOUS-WORK-QUEUE.txt. Read ENGINE-MAP.txt before code proposals and check MERGE-QUEUE.txt before duplicating work. Contributor task lists and handoff notes are stored under team/.

NEW AI JOINING? Do this:
1. Read AI-TEAM.txt, then TODO.txt (rules!), then your own list if you have one.
2. Add yourself under THE TEAM: name | role | folder | list.
   Pick a NEW folder with your name. Current active contributors must use their designated folder; Gemini now owns the consolidated gemini/ workspace.
3. Write a "JOINED" line in the WORK LOG below.
4. Only take jobs nobody has CLAIMED (see rules).

THE MAGIC WORD: "continue"
When the owner just says "continue" (and nothing else), it means:
  1. Re-read AI-TEAM.txt (work log), TODO.txt and your own list on GitHub.
  2. See what you last CLAIMED. If it isn't DONE, keep working on it.
     If it is DONE, CLAIM the next free job and start it.
  3. Reply first with one short line, e.g. "Got it, checking the to-do
     list... continuing with <job>." Then do the work.
  4. Don't ask the owner what to do, the files tell you. Only ask if
     something is truly unclear or would break a rule.
  5. If time remains after your task, claim the next suitable unclaimed task
     from team/CONTINUOUS-WORK-QUEUE.txt. Never invent busywork.

WHEN YOUR ASSIGNED LIST IS EMPTY
Do not stop just because your first task is done if session time remains. Check team/CONTINUOUS-WORK-QUEUE.txt and team/FUTURE-IDEAS.txt, then claim the next unclaimed, in-scope task matching your role. Review other AI work when assigned, but report findings instead of editing their folders. Make A/B/C variants only when explicitly requested. Claude Opus 5.5 alone handles official connected-runtime comparison/integration after model verification and owner authorization.
TOPICS (most important first; 60 FPS is an aspiration, not a guarantee):
  1. Scratch 3 behavior parity, including procedures/custom blocks, argument,
     recursion, variable scope, waits/yields, clones, events and save/load.
  2. Performance: target 60 FPS or better on representative projects and
     hardware when feasible; benchmark before/after instead of promising every
     heavy project will reach 60 FPS. Focus on script throughput, rendering,
     clones, collision, pen, lists, sounds and large costumes.
  3. A clean, simple primary UI; fix usability/blocking issues before adding clutter.
  4. Costumes / paint editor and costume animation.
  5. Original G-Warp visuals, icons, music and learning ideas.
  6. GWrap AI knowledge and safe in-app assistance.
The owner tests everything with screenshots, slowly, one by one.
If it works and runs properly, that's what matters.

GWRAP AI (the built-in assistant) - knowledge it needs
GWrap AI is an ASSISTANT inside the app, NOT a free-roaming autonomous AI.
It only helps with G-Warp: understanding the user's project, blocks,
sprites, costumes, sounds, extensions, fixing bugs, explaining ideas, and
doing what the user asks inside the app. Nothing beyond that.
Any AI can research and write its knowledge: how to read a project
(project.json), recognise every block and what it does, common bugs and
fixes, game recipes (platformer, shooter, clicker...), and safe answers
for kids. Put it in <your folder>/AI-KNOWLEDGE-<topic>.txt.
Never let it: send data without permission, run JavaScript blocks, act
outside G-Warp, or pretend to be a person.

CROSS-REVIEW RULE
Every AI may inspect and review any other AI's submitted work for bugs, missing requirements, unsafe assumptions, incomplete testing, and improvement ideas. Do not edit another AI's folder or silently rewrite their files. Submit a review with severity, exact file/function, evidence class, minimum correction, and acceptance test. The original author revises their own proposal; ChatGPT maintains the independent review ledger; Claude owns official integration evidence after model verification. A review is not permission to activate code.

TEAM RULES (to stop two AIs making the same thing)
1. CLAIM BEFORE YOU START: before working on a job, add a line to the
   WORK LOG:   CLAIMED | <your name> | <job>
   If someone already CLAIMED that job and it isn't DONE, pick another one.
2. When finished, add:  DONE | <your name> | <job> | <files you made>
3. If you find another AI already made the same thing, DON'T redo it.
   Write: REVIEWED | <your name> | <job> | <what could be better>
4. Only ever ADD lines at the bottom of the WORK LOG. Never delete or
   change another AI's lines.
5. Keep proposals in your own folder by default. If explicitly assigned to place code in app.html/index.html, add only a uniquely marked, inert, non-executing proposal block; never connect it to live behavior. If safe inert placement is uncertain, keep the full candidate in your own folder. Never touch another AI's folder.
6. All rules in TODO.txt still apply (same same but different, theirs is
   theirs, safety first, no cats).
7. READ ENGINE-MAP.txt BEFORE WRITING CODE. G-Warp is NOT scratch-vm:
   no window.vm, no runtime.sequencer. Code must use window.Engine and
   the GW_ names in ENGINE-MAP.txt, or say "UNSURE:".
8. NO PLACEHOLDERS: don't hand in empty/placeholder files or tiny stubs
   and call them DONE. If something is missing, log it as PARTIAL, not DONE.
9. GIT COMMIT MESSAGES must start with your name: "Grok: ...",
   "ChatGPT: ...". When preserving inherited lab work, say so explicitly, e.g.
   "Gemini (inherited lab): ..."; do not imply the original author committed it.
10. WORK LOG lines go at the very BOTTOM, in time order. Use only:
   JOINED, CLAIMED, PROGRESS, PARTIAL, DONE, REVIEWED, UNCLAIMED, NEW ROLE.
11. Don't commit junk files (.claude/, node_modules/, .DS_Store,
   scratch files). Only your folder + the shared .txt files.
12. Every code file needs a matching NOTES-<name>.txt or LAB-<name>.txt
   saying: what it does, where it plugs in, how to test it.
13. FOLLOW STYLE-GUIDE.txt: work goes INTO G-Warp's existing screens -
   never build a separate app/editor with its own name. Use G-Warp's
   dark electric colours.
14. NO OTHER PEOPLE'S CHARACTERS OR BRANDS - not even for testing:
   no Mario/Nintendo, Sonic, Pokemon, Minecraft, Disney, Scratch Cat...
   Use G-Spark or invented characters.

=====================================================================
 WORK LOG  (newest at the bottom - format: DATE | WHAT | WHO | JOB | FILES)
=====================================================================
2026-10-13 | DONE    | Claude  | website + Beta (v15.45.0) | index.html, app.html
2026-10-13 | DONE    | Claude  | team setup | TODO.txt, TODO-GROK.txt, AI-TEAM.txt
2026-10-13 | CLAIMED | ChatGPT | TODO.txt NOW 1 (freeze fixes) + NOW 2 (#scratch link)
2026-10-13 | CLAIMED | Grok    | TODO-GROK.txt G1-G5 (Scratch spec, tests, formats, extension check)
2026-10-13 | CLAIMED | Claude  | final merge of v15.46 + browser extension (after the others hand in)
2026-10-13 | PROGRESS| Grok    | G1 Scratch specs | SPEC-motion, SPEC-operators, SPEC-variables-lists, SPEC-control, SPEC-looks, SPEC-events, SPEC-sensing (still need sound, myblocks, pen, music)
2026-10-13 | DONE    | Grok    | G1 Scratch behaviour specs (all 11 categories) | grok/SPEC-motion.txt, SPEC-looks.txt, SPEC-sound.txt, SPEC-events.txt, SPEC-control.txt, SPEC-sensing.txt, SPEC-operators.txt, SPEC-variables-lists.txt, SPEC-myblocks.txt, SPEC-pen.txt, SPEC-music.txt
2026-10-13 | JOINED  | DeepSeek | researcher, no GitHub (owner hands work over)
2026-10-13 | CLAIMED | DeepSeek | TODO.txt NOW 5 (.pmp files open)
2026-10-13 | JOINED  | Gemini   | researcher, no GitHub (owner hands work over)
2026-10-13 | DONE    | Gemini   | first draft of the browser extension (saved by Claude; needs fixes)
2026-10-13 | CLAIMED | Gemini   | TODO.txt NOW 3 (browser extension, fixed version)
2026-10-13 | DONE    | Gemini   | TODO.txt NOW 3 (browser extension, fixed) | gemini/extension/ (files saved to GitHub by Grok on owner handoff; Claude still needs to test + merge; PNG icons via make-icons.js; gspark.svg is a placeholder bolt)
2026-10-13 | DONE    | DeepSeek | TODO.txt NOW 5 research (partial) | deepseek/NOTES-pmp.txt, deepseek/test-pmp-loader.js (pmp-loader.js MISSING from paste — see deepseek/MISSING-pmp-loader.txt; Claude does not merge until loader arrives)
2026-10-13 | UNCLAIMED| Gemini  | TODO.txt NOW 3 (browser extension) - free for another AI; Claude will finish it if nobody takes it
2026-10-13 | NEW ROLE | Gemini  | icons + pictures (3 SVG options each) -> TODO-GEMINI.txt
2026-10-13 | NEW ROLE | DeepSeek | lab scientist: 3+ versions (A/B/C) per job -> TODO-DEEPSEEK.txt
2026-10-13 | CLAIMED | Grok    | TODO-GROK.txt G2 (TESTS.txt 60+ Scratch edge-case scripts)
2026-10-13 | DONE    | Grok    | TODO-GROK.txt G2 | grok/TESTS.txt (75 cases T001-T075)
2026-10-13 | DONE    | Grok    | TODO-GROK.txt G3 | grok/SPEC-engine.txt
2026-10-13 | DONE    | DeepSeek | 3-code tester: SPEED (60 FPS) | deepseek/speed-collision-A/B/C.js, speed-clones-A/B/C.js, speed-pen-A/B/C.js, LAB-speed.txt (saved by Grok; Claude must test before merge; recommended combo collision-B + clones-C + pen-A)
2026-10-13 | REVIEWED | Claude  | team check: app.html/index.html untouched (good); folders respected (good); Grok's specs+tests good. Problems: DeepSeek speed code targets window.vm/runtime.sequencer which G-Warp doesn't have (won't run) -> added ENGINE-MAP.txt + rule 7; pmp-loader.js still missing; Gemini gspark.svg is a placeholder; ChatGPT nothing handed in yet. New rules 7-12 + .gitignore.
2026-10-13 | UNCLAIMED| DeepSeek | speed A/B/C must be REDONE against window.Engine (see ENGINE-MAP.txt) - DeepSeek's job
2026-10-13 | DONE    | DeepSeek | speed A/B/C REDONE for window.Engine + pmp-loader.js | deepseek/speed-*-[ABC].js, pmp-loader.js, LAB-speed.txt (saved by Claude; MISSING-pmp-loader.txt removed)
2026-10-13 | REVIEWED | Claude  | DeepSeek speed v2: well made + safe (bails out if names missing) BUT the guessed names (Engine.touching, stepThread, Engine.pen) don't exist, so all 9 would do nothing. Real hooks now listed in ENGINE-MAP.txt (GW.spritesTouch, ScriptThread.step, pen needs a hook). pmp-loader: ZIP part duplicates what G-Warp already does; only the PROTOBUF part is new and useful.
2026-10-13 | DONE    | Grok    | TODO-GROK.txt G4 | grok/SPEC-files.txt
2026-10-13 | DONE    | DeepSeek | speed v3 (real names: GW.spritesTouch, ScriptThread.step) + protobuf-only pmp-loader | deepseek/speed-collision-[ABC].js, speed-clones-[ABC].js, pmp-loader.js, LAB-speed.txt (pen files removed: no pen API) - saved by Claude
2026-10-13 | REVIEWED | Claude  | v3: collision A/B/C now hook the REAL function -> usable, B is the pick. Clones A/B/C: ScriptThread.blocks is the whole script (never empty) so "empty stack" never happens -> they skip nothing; also invisible clones still have to run their scripts. Better idea: skip threads that are waiting (waitUntil). Real ScriptThread fields added to ENGINE-MAP.txt.
2026-10-13 | CLAIMED | ChatGPT | URGENT U1 (UI 1-to-1 research) - assigned by owner
2026-10-13 | CLAIMED | DeepSeek | URGENT U2 (costume / paint editor fixes) - assigned by owner
2026-10-13 | CLAIMED | Claude  | URGENT U3 (bug hunt, clean-up, merge, next Beta)
2026-10-13 | DONE    | DeepSeek | URGENT U2 costume editor A/B/C (select, fill, text, eraser, undo) + checklist | deepseek/costume-*-[ABC].js, LAB-costumes.txt, CHECKLIST-costumes.txt (saved by Claude)
2026-10-13 | REVIEWED | Claude  | U2: big, careful work; all 15 files pass a syntax check; research + 40-point checklist are useful. But it guesses state.mode / fillColor / eraserSize / font - real fields are costumeMode, svgTool, bitmapTool, selectedSvg (added to ENGINE-MAP.txt). Claude will adapt the picks (select-C, fill-C, text-B, eraser-C, undo-C) when merging.
2026-10-13 | DONE    | ChatGPT | URGENT U1 UI behavior audit (static source audit; browser verification left for U3) | chatgpt/UI-AUDIT.txt
2026-10-13 | REVIEWED | Claude  | tidy-up: ChatGPT's clone-safety.js + scratch-link.js pass syntax check and use real names (good); UI-AUDIT is thorough. Added MERGE-QUEUE.txt (what's ready / reference / superseded), [~] status in TODO.txt, marked U2 done.
2026-10-13 | DONE    | Claude  | docs: README.md expanded (features, files, team, roadmap, bug reports, credits) + STYLE-GUIDE.txt + rules 13-14 (no separate apps; no Mario/Nintendo or other brands/characters, even for tests)
2026-10-13 | DONE    | ChatGPT | Screenshot/UI behavior documentation + complete user guide + two schematic diagrams + docs index (not browser screenshots) | chatgpt/SCREENSHOT-UI-DOCUMENTATION.md, chatgpt/USER-GUIDE.txt, chatgpt/images/*.svg, chatgpt/README.md
2026-10-13 | DONE    | Claude  | ChatGPT's request: live UI verification - 11 real screenshots (docs/screens/), 10 live tests all PASS, report docs/LIVE-UI-VERIFICATION.txt, real screenshots added to chatgpt/SCREENSHOT-UI-DOCUMENTATION.md. UI-AUDIT #1 (category single click) NOT confirmed - it works live.
2026-10-13 | DONE    | ChatGPT | organized docs: central index, reconciled live evidence, fixed stale category audit, user guide/navigation cleanup; Cloud AI note marked superseded | docs/README.md, README.md, chatgpt/README.md, chatgpt/README.txt, chatgpt/SCREENSHOT-UI-DOCUMENTATION.md, chatgpt/USER-GUIDE.txt, chatgpt/UI-AUDIT.txt, chatgpt/CLOUD-AI-LIVE-VERIFICATION-NOTE.txt
2026-10-13 | DONE    | ChatGPT | rechecked documentation goal; added beginner learn-by-doing guide and deep Scratch/G-Warp engineering reference, reorganized central index, linked detailed guides to both paths; Scratch architecture references checked | docs/LEARNING-GUIDE.md, docs/ENGINEERING-REFERENCE.md, docs/README.md, chatgpt/USER-GUIDE.txt, chatgpt/SCREENSHOT-UI-DOCUMENTATION.md
2026-10-13 | DONE    | ChatGPT | added three explanatory SVG diagrams and embedded them in beginner guide, engineering reference, and docs index | docs/images/gwarp-editor-map.svg, docs/images/gwarp-blocks-explained.svg, docs/images/gwarp-scratch-architecture.svg
2026-10-13 | DONE    | ChatGPT | embedded SVG illustrations inline inside the learning guide, engineering reference, central index, detailed user manual, and UI documentation (not merely linked as separate files) | docs/LEARNING-GUIDE.md, docs/ENGINEERING-REFERENCE.md, docs/README.md, chatgpt/USER-GUIDE.txt, chatgpt/SCREENSHOT-UI-DOCUMENTATION.md
2026-10-13 | CLAIMED | ChatGPT | documentation gap audit: expand research and engineering library | docs/RESEARCH-LIBRARY.md, docs/RUNTIME-SEMANTICS-AND-PARITY.md, docs/PROJECT-FORMATS-AND-ROUNDTRIP.md, docs/EXTENSIONS-SECURITY-MODEL.md, docs/COMPATIBILITY-TEST-MATRIX.md, docs/EDITOR-ENGINEERING-PLAYBOOK.md
2026-10-13 | DONE | ChatGPT | documentation gap audit: expanded source-linked research, runtime semantics, file formats, extension security, editor design and compatibility verification; added embedded diagrams and updated main documentation navigation | docs/README.md, README.md, chatgpt/README.md, docs/*.md, docs/images/compatibility-test-pipeline.svg, docs/images/extension-risk-boundaries.svg
2026-10-13 | REVIEWED | ChatGPT | corrected documentation direction: removed generic research/checklist documents and replaced the main reference with a source-grounded G-Warp implementation map and application diagrams | docs/ENGINEERING-REFERENCE.md, docs/README.md, README.md, chatgpt/README.md, docs/images/gwarp-app-internals.svg, docs/images/scratch-to-gwarp-map.svg
2026-10-14 | DONE | ChatGPT | master AI workflow plan, clarified roles/model gate, and moved task/handoff files under team/; app.html/index.html untouched | team/AI-WORK-PLAN.txt, team/README.txt, team/TODO-GROK.txt, team/TODO-GEMINI.txt, team/TODO-DEEPSEEK.txt, team/CLOUD-AI-OPUS-5.5-HANDOFF.txt
2026-10-14 | DONE | ChatGPT | added universal dormant-code proposal rule, Claude-only connection/visual comparison workflow, and owner-controlled future ideas parking lot | team/AI-WORK-PLAN.txt, team/CLOUD-AI-OPUS-5.5-HANDOFF.txt, team/FUTURE-IDEAS.txt

2026-10-15 | NEW ROLE | ChatGPT | standing independent reviewer of every AI's deliverables; maintain review ledger and issue bounded fixes/tests | team/TODO-CHATGPT.txt, team/CROSS-AI-REVIEW-2026-10-15.txt
2026-10-15 | DONE | ChatGPT | first-pass static cross-AI audit; identified costume API/canvas risks, clone-safety validation needs, incomplete PMP assets/fixtures, Grok semantic uncertainty and Gemini packaging/browser gaps | team/CROSS-AI-REVIEW-2026-10-15.txt
2026-10-15 | DONE | ChatGPT | created continuous bounded task queue, added cross-review duties to all contributor task lists, and defined Claude integration/evidence-owner mode | team/CONTINUOUS-WORK-QUEUE.txt, team/AI-WORK-PLAN.txt, team/TODO-GROK.txt, team/TODO-DEEPSEEK.txt, team/TODO-GEMINI.txt
2026-10-15 | PARTIAL | ChatGPT | security-pattern scan + consistency audit; no matching hard-coded patterns found in current scanned files; history not scanned; canonical website URL awaiting owner | team/SECURITY-AND-CONSISTENCY-AUDIT-2026-10-15.txt, team/CROSS-AI-REVIEW-2026-10-15.txt, team/FUTURE-IDEAS.txt, TODO.txt, README.md

 
FORMER DEEPSEEK ROLE — RETIRED / CONSOLIDATED (2026-10-18)
The owner reassigned DeepSeek's lab/research/candidate duties to Gemini 2 (accessed through FreeBuff/Argon). Active work, new candidates, test harnesses and reports now belong under gemini/. Legacy files were moved to gemini/lab/ and must be reviewed before reuse; their presence does not prove correctness or integration. Do not create new deepseek/ files, use the temporary-coverage rule, or treat the retired return-handoff template as active. Existing DeepSeek work-log entries remain historical evidence. Claude's integration/model-verification gates and owner approval are unchanged.


SECURITY RULE — API KEYS AND SECRETS (2026-10-15)
Never commit or paste real API keys, access tokens, passwords, private keys, cookies, or credentials into source, screenshots, TXT reports, tests, issues, prompts, or handoffs. Use dummy placeholders only. Never ask the owner to paste a live key into chat. Never embed a shared provider key in app.html, index.html, an extension, or any public browser bundle. User-owned keys must have clear provider/destination/cost/context disclosure; session-only storage should be the default and persistent browser storage must warn that it is not a secure vault. Never include keys in exports, project data, logs, diagnostics, or reports. If a real key may be exposed, do not reproduce it; alert the owner by path/line and advise revocation/rotation and history review. A public browser app cannot guarantee that a shared secret is unextractable. Do not create a backend or change key behavior without owner approval. See team/SECURITY-AND-CONSISTENCY-AUDIT-2026-10-15.txt.

2026-10-16 | REVIEWED | ChatGPT | main-branch integration/status audit: app.html/index.html blob SHAs unchanged from prior baseline; no new app integration detected; clone/scratch-link candidates remain disconnected; Gemini extension is separate prototype with packaging/browser-test blockers; old candidate branches are diverged | team/INTEGRATION-STATUS-AUDIT-2026-10-16.txt

2026-10-16 | REVIEWED | ChatGPT | Gemini-only static review: current main has no newer Gemini branch/commit; extension is separate from G-Warp runtime; PNG icons missing; SPA wrapper, privacy copy, storage errors, URL confirmation and browser/end-to-end tests remain | team/GEMINI-ONLY-REVIEW-2026-10-16.txt
\n2026-10-17 | RULE ADDED | ChatGPT | Added mandatory per-AI AI-CHAT-LOG.txt request/response preservation rule; transcript must not fabricate missing history or store secrets | team/AI-CHAT-LOG-TEMPLATE.txt\n
2026-10-17 | RULE ADDED | ChatGPT | Required per-AI AI-CHAT-LOG.txt to preserve user requests and AI responses when chat history is truncated; template created | team/AI-CHAT-LOG-TEMPLATE.txt
2026-10-17 | PLAN    | Claude  | read-only check: app.html/index.html unchanged since Beta v15.45.0; plan written -> team/CLAUDE-PLAN-2026-10-17.txt (waiting for owner's go)


PROGRESS / PLAN UPDATE (2026-10-18)
- Added team/AI-PROGRESS-PREVIEW-2026-10-18.txt: concise status of ChatGPT, Grok, DeepSeek, Gemini, Buffy-labelled candidates and Claude, with evidence levels and blockers.
- Added team/CLAUDE-EXECUTION-PLAN-2026-10-18.txt: phased integration plan prioritizing baseline verification and freeze/Stop reproduction, with token-saving workflow, test gates, rollback and owner approval.
- Linked both documents from team/AI-WORK-PLAN.txt. These are planning/reporting changes only; no app.html/index.html code was changed and no new browser tests were run.


2026-10-18 | DONE | Grok | Restored AI-TEAM.txt after accidental truncate during G5 log append | source commit 8f5d9a4ca69db8b1f314cec683e628883177292c; prior blob c0ee70721157421b6fcf8b2183e2c1d06e31cbd4

2026-10-18 | DONE | Grok | G5 extension gallery audit | grok/EXTENSIONS-check.txt (184 cards; 17 LIST matched; no unmarked personal-data hits in static scan; 32 URLs HTTP OK; app.html untouched)
2026-10-18 | DONE | ChatGPT | Simplified repository workflow routing; restored/verified AI-TEAM log, created single entry-point guide, superseded prior Claude plan, and started duplicate-document cleanup | team/WORKFLOW-START-HERE.txt

2026-10-18 | DONE | ChatGPT | Completed workflow cleanup: canonical workflow/status/checkpoint, shorter roadmap/queue, archived duplicate snapshots and old Claude instructions, removed duplicate ChatGPT README; app.html/index.html unchanged and no runtime tests | team/WORKFLOW-START-HERE.txt; team/AI-STATUS.txt
2026-10-18 | PROGRESS | Claude | P0 clone freeze reproduced + fixed in a v15.46.0 candidate (not uploaded); zombie.sb3 tested; report team/P0-CLONE-FREEZE-REPORT.txt; checkpoint updated
2026-10-18 | DONE    | Claude | v15.46.0 candidate: chaos test ALL PASS, regression 432/438 (6 = title-only checks) | claude/app-v15.46.0-candidate.html (waiting for owner OK before replacing app.html)
2026-10-18 | RELEASED | Claude | Beta v15.46.0 -> app.html (owner said push): freeze fixes (O(1) clone count, 5000 ceiling for infinite clones, frame budget, watchdog with 6 s start grace, text/list caps, crash recovery), faster sound restarts + computed menus, new More -> Report a bug (opens a pre-filled GitHub issue the user sends; errors kept locally only). Tests: chaos ALL PASS, bug report ALL PASS, regression 432/438 (6 title-only)
2026-10-18 | RELEASED | Claude | Beta v15.46.1: faster "item # of" (2x, same answers in 13,200 random checks); key tests pass; speed plan team/SPEED-PLAN.txt (statement cache -> compiler next)
2026-10-18 | MERGED  | Claude | chatgpt/scratch-link.js into app.html (Beta v15.46.1): #scratch=<id> asks, then loads; Cancel/bad id do nothing (test_4611 ALL PASS). Gemini's extension link target now works.
2026-10-18 | ASSIGNED | Claude | jobs for every AI -> team/ASSIGNMENTS-2026-10-19.txt (Grok GR1-2, DeepSeek DS6-7, Gemini GM7-8, ChatGPT CG1-3, Claude CL1-3)
2026-10-18 | RELEASED | Claude | Beta v15.46.2: skip inaudible sound restarts (<30 ms); key tests pass. Found zombie game runs 3.2M if/else + 3.2M start sound in 15 s -> compiler (CL1) is the real fix, see team/SPEED-PLAN.txt
2026-10-18 | RELEASED | Claude | Beta v15.46.3 COSTUMES: Paint new makes a vector costume (was bitmap, Scratch makes vector); vector RESHAPE + ERASER work (were 'coming soon') - claude/vectools4630.js, built on DeepSeek's checklist/lab. Costume checklist test claude/tests/test_cos.js 20/20 PASS + old costume tests 3300/3700/3900/4000/2400 PASS
2026-10-18 | RELEASED | Claude | Beta v15.46.4 POWER LINES (owner's animation idea): while a project runs, ghosted electric wires show broadcast -> when I receive (green) and My Block -> define (violet), sparks travel on each message, start blocks get a sparking pole, broadcasts heard by other sprites name them; switch in More -> Settings. Drawn outside the code canvas (no cache churn), clicks pass through, Stop clears it. claude/powerlines4640.js; test claude/tests/test_4640.js ALL PASS; cos/4600/4601/4611/4700 still ALL PASS; zombie FPS unchanged (layout cost ~0.6 ms per half second)
2026-10-18 | PROGRESS | ChatGPT | CG2 initial static review of Claude's speed/compiler plan; confirmed clone-safety candidate already marked SUPERSEDED; statement-cache/compiler code not yet present in inspected history | chatgpt/COMPILER-REVIEW-2026-10-18.txt
2026-10-18 | RELEASED | Claude | Beta v15.47.0 COSTUME ANIMATION (owner's focus): Flipbook in the Costumes tab (Onion skin: costume before violet / after green behind the drawing, lined up by rotation centre, vector+bitmap, never saved; Play: flips the costumes, FPS/loop/range). Animation tab (bones) + Animation blocks out of Developer-mode preview, for everyone; G-Spark opens with ready-made bones + walk/wave; costumes made from an animation follow its changes (rev stamp), rename renames costumes + block menus, <- -> and Space keys. claude/flipbook4700.js; test claude/tests/test_cosanim.js 27/27 PASS; cos/4600/4601/4611/4700/4640 still ALL PASS; old anim tests only differ where the button/blocks are now shown to everyone (intended)
2026-10-18 | ASSIGNED | Claude | costume-animation jobs -> team/ASSIGNMENTS-ANIMATION-2026-10-18.txt (Grok GA1-2, DeepSeek DA1-3, Gemini GA-M1-2, ChatGPT CA1-3, Claude CL4-6); TODO-<AI>.txt headers point there first
2026-10-18 | SETUP | Claude | owner's 5 test projects packed for the team (each with SUMMARY.txt, SCRIPTS.txt = every script as text, v15.47.0 baseline) for the PRIVATE repo Creapter1/gwarp-test-projects (owner creates it, Claude uploads). Rule 7 in TODO.txt: never copy them into public places. Baselines: grass 60 FPS, space game 60 FPS, plaing game 51 FPS, zombie 2 FPS (speed target), The worst day 23 FPS / 10 s load; found: 'item # of' and '<' marked unsupported in The worst day (to fix). claude/tests/smoke.js = the baseline runner

2026-10-18 | NEW ROLE | Gemini 2 (FreeBuff/Argon) | inherits all active DeepSeek lab/research/A-B-C/test-harness duties; one unified workspace gemini/; DeepSeek role retired by owner | team/TODO-GEMINI.txt; gemini/lab/
2026-10-18 | DONE | ChatGPT | owner-directed Gemini/DeepSeek role consolidation; migrated 28 legacy lab files to gemini/lab/, retired active DeepSeek routing, preserved owner wording + clone provenance review, added performance benchmark protocol; app.html/index.html untouched, no runtime tests | team/TEAM-REASSIGNMENT-2026-10-18.txt; team/TODO-GEMINI.txt; team/PERFORMANCE-BENCHMARK-PLAN.txt; chatgpt/MERGE-QUEUE-NOTES.txt; gemini/lab/
2026-10-19 | REVIEWED | Claude | Grok GA1 (20/20 PASS in G-Warp), GA2 (accepted), GR1/GR2 (26/30 first run: 3 real G-Warp bugs fixed, T043 is wrong - repeat uses Math.round, see claude/REVIEW-GROK-2026-10-19.txt). Runners: claude/tests/test_grok_anim.js, test_grok_fixtures.js (CL3 done)
2026-10-19 | RELEASED | Claude | Beta v15.48.0 COSTUME EDITOR EMERGENCY FIX (owner: letters and shapes jump; no Scratch handles): new Select tool with 8 handles + rotate (claude/transform4800.js), nothing jumps (text, shapes, brush, imported costumes, after flip/rotate), multi-select, group/ungroup, Paint new fits the stage-size area and its drawing, costumes grow instead of cutting drawings off, Convert to Bitmap keeps size, Bitmap Select moves/resizes pixels (no dashed lines saved), circle reshape no longer tears, brush/eraser never drag shapes. Found by a QA agent run (report: scratchpad costume-qa/REPORT.txt), verified by claude/tests/test_select.js (all pass) + old costume tests. Also speed: zombie test game 2 -> 58 FPS (is-sound-playing with a joined name was lost on import -> endless loop), The worst day 23 -> 37 FPS; set drag mode + Clones+ touching-clone-with-var blocks; Scratch parity fixes for letter of / list positions / number display
2026-10-19 | UPDATED | Claude | Beta v15.48.0 (same version): a shape just drawn with Rectangle/Circle/Line keeps its handles like Scratch (resize it right away). Owner out of usage -> other AIs: check team/NEXT-FOR-THE-TEAM-2026-10-19.txt
2026-10-19 | TESTED | Claude | full regression of Beta v15.48.0: 429 PASS, 9 FAIL - all expected (old version-number checks, Animation now shown to everyone, Reshape/Eraser now enabled). No new problems.
2026-10-19 | ASSIGNED | Claude | Gemini 3 (Argon) big list: 7 picture jobs, 8 lab jobs (3 versions each), 3 writing jobs -> team/GEMINI-BIG-LIST-2026-10-19.txt
2026-10-19 | REVIEWED | ChatGPT | repository-wide source/tree verification: role migration and 28-file move confirmed (25 blobs unchanged, 3 docs intentionally updated); current Beta v15.48.0 regression is Claude-reported 429 PASS / 9 expected FAIL; Grok corrected T043 in fixtures.json but source map and 60+10 runner integration remain open; Gemini 2/3 Argon label mismatch logged; no runtime test by ChatGPT | chatgpt/VERIFICATION-2026-10-19.txt; chatgpt/MERGE-QUEUE-NOTES.txt; team/AI-STATUS.txt
2026-10-19 | REVIEWED | ChatGPT | follow-up verification: Grok corrected T043 in fixtures.json (60 base cases), but source map still says 2 iterations and current 60+10 list runner is not wired; Claude logs full v15.48.0 regression 429 PASS / 9 expected FAIL; reconciled Gemini Big List vs older animation tasks to prevent duplicates; flagged Gemini 2/3 name and canonical URL conflicts; no runtime test by ChatGPT | chatgpt/VERIFICATION-2026-10-19.txt; team/TODO-GEMINI.txt; team/AI-STATUS.txt
2026-10-19 | REVIEWED | ChatGPT | final consistency pass: current app/index hashes unchanged by ChatGPT; 28 legacy DeepSeek files preserved under gemini/lab/ (25 byte-identical, 3 documented edits); Grok T043 corrected in 60-case JSON but source map stale and T130-T139 not wired into current runner; Gemini Big List mapped to older overlapping assignments; name/URL conflicts documented; runtime tests not run by ChatGPT | chatgpt/VERIFICATION-2026-10-19.txt; team/TODO-GEMINI.txt; team/AI-STATUS.txt
2026-10-19 | ASSIGNED | Claude | ChatGPT: website update + full documentation (14 docs pages) -> team/CHATGPT-WEBSITE-DOCS-2026-10-19.txt

2026-10-19 | ChatGPT WEBSITE/DOCS PASS 1
- Added 14 responsive documentation pages under docs/ and linked the docs home from index.html. Homepage now has a Beta v15.48.0 highlights section and 10 FAQ cards.
- Added chatgpt/WEBSITE-OPEN-QUESTIONS.txt to keep unverified current-app behavior, canonical URL, offline/privacy claims, shortcut details and credits scope explicit.
- Files: index.html; docs/index.html, start.html, blocks.html, costumes.html, animation.html, sounds.html, extensions.html, safety.html, credits.html, scratch.html, troubleshoot.html, shortcuts.html, changelog.html, team.html; chatgpt/WEBSITE-OPEN-QUESTIONS.txt; chatgpt/AI-CHAT-LOG.txt.
- Evidence: static GitHub file/link check only. No browser/runtime/published-site test. app.html untouched. Do not mark the full website assignment complete until link check and live-site review are done.
2026-10-19 | ASSIGNED | Claude -> Grok | big job list team/GROK-BIG-LIST-2026-10-19.txt (R0-R8 fixtures, S1-S4 specs, V1-V2 reviews)
2026-10-18 | PARTIAL | ChatGPT | website/docs pass 2: filled seven missing feature cards, removed stale fixed extension count, improved Stage/backdrop, comments/Bug Hunter, clone-safety/recovery and Look-like-Scratch docs; current-main static link check 0 missing local targets; confirmed root MIT LICENSE. No app.html edit and no browser/published-site test; extension count/native catalog and four Animation block names remain unverified | index.html; docs/start.html; docs/blocks.html; docs/troubleshoot.html; docs/scratch.html; chatgpt/WEBSITE-OPEN-QUESTIONS.txt; chatgpt/AI-CHAT-LOG.txt
2026-10-19 | ASSIGNED | Claude -> ChatGPT round 2 (tutorials, glossary, parents, blocks upgrade, README, a11y) + Grok extra (review tutorials)
2026-10-18 | PARTIAL | ChatGPT | website/docs pass 3 continued: added three start actions and five missing v15.46–v15.48 highlight cards; removed stale extension-count infographic, softened unverified parity/safety claims, expanded Power lines/Paint new docs, split changelog by release, and added AI-tool/MIT credits. Rechecked all 15 public HTML pages: 339 local href/src references, zero missing file targets; all eight new card target anchors exist. No browser/live-site/mobile verification; count, screenshot freshness and four Animation block names remain pending | index.html; docs/blocks.html; docs/costumes.html; docs/changelog.html; docs/credits.html; chatgpt/WEBSITE-OPEN-QUESTIONS.txt; chatgpt/AI-CHAT-LOG.txt
2026-10-19 | ASSIGNED | Claude -> Gemini | MELODY job M1-M2 (top priority), owner says all modes sound the same
2026-10-18 | PARTIAL CONTENT / STATIC CHECKS | ChatGPT | Round 2 D1–D6: added eight tutorials plus hub, 65-term glossary, parent/teacher privacy/classroom guide, 100+ block-reference examples, concise README, and visible keyboard-focus styling across 25 documentation pages. Static scan found 0 missing local href/src targets, duplicate IDs, heading-level skips, or missing image alt attributes (no inline images). Browser, screen-reader, runtime, exact palette-label, extension inventory and published-site claims remain unverified; Claude owns the browser/mobile pass. app.html unchanged | docs/tutorials/*.html; docs/glossary.html; docs/parents.html; docs/blocks.html; docs/index.html; README.md; remaining docs HTML; chatgpt/WEBSITE-OPEN-QUESTIONS.txt; chatgpt/AI-CHAT-LOG.txt
2026-10-18 | PARTIAL / STATIC REVIEW | ChatGPT | Completed costume-animation review tasks CA1–CA3: added “Animate your costumes” to chatgpt/USER-GUIDE.txt; created chatgpt/REVIEW-costume-animation.txt with 14 untested coverage cases based on claude/tests/test_cosanim.js; created chatgpt/UX-costume-animation.txt with 10 ranked UX recommendations and source links. No Playwright run, no app execution, and no edits to app.html or Claude's test file | chatgpt/USER-GUIDE.txt; chatgpt/REVIEW-costume-animation.txt; chatgpt/UX-costume-animation.txt; chatgpt/AI-CHAT-LOG.txt
2026-10-19 | PARTIAL REVIEW / DOCS | ChatGPT | CG1–CG3: updated chatgpt/USER-GUIDE.txt with bug-report, clone-cap, stuck-project and #scratch= steps; created chatgpt/COMPILER-REVIEW-2026-10-19.txt identifying no new statement-cache/compiler implementation and the GW.__wsEpoch + fixture coverage acceptance risks; appended queue refresh recommendations to chatgpt/MERGE-QUEUE-NOTES.txt without editing Claude-owned MERGE-QUEUE.txt. Static only; no app/test code edits or browser run. Next C1 Gemini costume-candidate review | chatgpt/USER-GUIDE.txt; chatgpt/COMPILER-REVIEW-2026-10-19.txt; chatgpt/MERGE-QUEUE-NOTES.txt; chatgpt/AI-CHAT-LOG.txt
2026-10-19 | PARTIAL / STATIC REVIEW | ChatGPT | C1: created chatgpt/REVIEW-GEMINI-COSTUME-CANDIDATES-2026-10-19.txt; all five inherited costume candidates blocked/high risk pending corrected v15.48 API mapping, canvas-scoped events, model persistence and undo tests. Did not edit gemini/lab/ or app.html. Next C2 clone-safety static review | chatgpt/REVIEW-GEMINI-COSTUME-CANDIDATES-2026-10-19.txt; chatgpt/AI-CHAT-LOG.txt
2026-10-19 | PARTIAL / STATIC REVIEW | ChatGPT | C2: created chatgpt/REVIEW-clone-safety-2026-10-19.txt with side-effect analysis and P0/P1 regression checklist. Candidate chatgpt/clone-safety.js remains SUPERSEDED per current queue; no revival, no code/app edits, no tests run. Findings include absent 5,000 ceiling in Infinite Clones, Stop-thread ordering risk, native counter-cache bookkeeping divergence, and project clone-limit compatibility risk | chatgpt/REVIEW-clone-safety-2026-10-19.txt; chatgpt/AI-CHAT-LOG.txt
2026-10-19 | PARTIAL / STATIC REVIEW | ChatGPT | C3: created chatgpt/REVIEW-SCRATCH-LINK-EXTENSION-2026-10-19.txt. App hash handler already integrated; extension still has package icon, preference sync, SPA/history cleanup, origin-validation, privacy wording and browser matrix open. No app/Gemini edits and no tests run. Next C4 Grok test expected-result evidence review | chatgpt/REVIEW-SCRATCH-LINK-EXTENSION-2026-10-19.txt; chatgpt/AI-CHAT-LOG.txt

2026-10-19 | CLAIMED | ChatGPT | C4 review Grok TESTS.txt expected results and primary-source requirements
2026-10-19 | DONE | ChatGPT | C4 static review -> chatgpt/REVIEW-GROK-EXPECTED-RESULTS-2026-10-19.txt (commit dc28e2df1b18927d1c1d2a9c448a98ff6f9b0ad9); Scratch editor VM source cross-checked at scratchfoundation/scratch-editor commit 585b1c3877943b31ac273b3d056ccb79db13aa41; T043 is still wrong in Grok TESTS.txt, T017 has invalid random-test logic, and UI/G-Warp-only cases need evidence/scope cleanup; no runtime test performed
2026-10-19 | CLAIMED | ChatGPT | C5 review gemini/lab/pmp-loader.js and test-pmp-loader.js (read-only; verify parser API, legal real fixture/assets, and test limits)

2026-10-19 | CLAIMED | ChatGPT | C5 review inherited protobuf .pmp loader/test harness (read-only)
2026-10-19 | DONE | ChatGPT | C5 static review -> chatgpt/REVIEW-PMP-PROTOBUF-LOADER-2026-10-19.txt (commit 61011ebf5ff473c23d0c1f8e6cd582c5b3f19094); current upstream pmp-protobuf API cross-checked at commit 84176d56c2b5bdebce8150a6bf9fefcac59fc1af; real protobuf fixture/assets and browser import remain untested, empty assets Map blocks full media load; Gemini files/app untouched
2026-10-19 | CLAIMED | ChatGPT | C7 concise owner report for C1-C5 review cycle; keep all runtime/integration work out of scope

2026-10-19 | CLAIMED | ChatGPT | C7 concise owner handoff report for the C1-C5 review cycle
2026-10-19 | DONE | ChatGPT | C7 owner report -> chatgpt/OWNER-REVIEW-STATUS-2026-10-19.txt (commit 6bf076af44099ec43579e191831a46c1e8c310a8); summarizes blocked/superseded candidate status, required Grok/Gemini/Claude actions and owner website-URL decision; no runtime test or integration

2026-10-19 | CLAIMED | ChatGPT | C8 audit user-owned AI key handling / exact project context and reconcile stale TODO statuses
2026-10-19 | DONE | ChatGPT | C8 static security audit -> chatgpt/REVIEW-AI-KEY-CONTEXT-SECURITY-2026-10-19.txt (commit f1265a8582f51a6049a323614f2ab9b7166b6a87); existing security audit addendum commit cd95b4b41563f3ecab1ec199f83663047daa4a21; root TODO updated to distinguish feature source presence from privacy/model acceptance and historical 59 case results from current 60+10; no browser/API/history scan or app edit

2026-10-19 | DONE | ChatGPT | C8 follow-up consistency check: corrected stale task routing in team/TODO-CHATGPT.txt (commit cc01d8d15b27053569219478f071ce1bec7a20f4); website/docs, CA1-CA3, and CG1-CG3 are prior delivered outputs, not current work; C6 ledger duty continues | team/TODO-CHATGPT.txt; chatgpt/AI-CHAT-LOG.txt
2026-10-19 | ASSIGNED | Claude -> Grok | BIG LIST 2 team/GROK-BIG-LIST-2-2026-10-19.txt (K0 fix TESTS T043/T017, K1-K4 cast+compiler+runner+design, K5 app bug hunt, K6-K7 coverage, K8-K10 music data + reviews, K11-K12 extension specs)
2026-10-19 | ASSIGNED | Claude -> Gemini | 35-min SPRINT team/GEMINI-SPRINT-2026-10-19.txt (melody engine first, then rig characters, flipbook sprites, easing, backdrops, empty states, touching bench)


2026-10-19 | DONE | ChatGPT | C9 static review of Gemini animation/lab batch b05973868839921586c844a93c4e1546c75c1c5c -> chatgpt/REVIEW-GEMINI-ANIMATION-LAB-2026-10-19.txt; current M1 paths/pivots.json absent, generic-root self-parent risk and incomplete assertions identified; no Node/browser test or app edit; C6 ledger updated.
2026-10-19 | RELEASED | Claude | Beta v15.49.0 COSTUME TABS + TEAM MERGE (owner's 1 a.m. list): Costume list shows folders (walk - 8, opens/closes, opens itself for the costume showing). Costume tabs on top of the costume section - Costumes | Tiles | Animation (claude/costabs4900.js; Animation is a real tab now). Tile sprites v1: Tiles only for an empty sprite, 'Turn this sprite into a tile map' turns Costumes + Animation off, can be turned back unless a costume was added (kept in autosave). Glide button in the Animation editor (3 eased in-betweens per gap, ease curve from Gemini's easing lab; press again = key frames). Animation costumes in TurboWarp-style folders: walk//frame 1..N (make, rename, blocks). MERGED Gemini's melody engine into the Song maker (each style its own melody) + new styles Chiptune/Rock/Trap/Cinematic. MERGED Gemini's 3 backdrops into Choose a Backdrop. Tests: test_cosanim 25/27 (the 2 fails also fail on v15.48.0: drawing a 2-shape costume in that test), test_select all pass, Tiles/Glide/melody/backdrop probes pass, full regression running.
2026-10-19 | ASSIGNED | Claude -> Gemini, Grok, ChatGPT | team/NEXT-AFTER-v15.49-2026-10-19.txt (Gemini: ground line lab, melody v2 backing, folders mockup, tiles art, icon picks; Grok: K0 then ground-line + folders specs then BIG LIST 2; ChatGPT: small docs updates only - no review passes). Ideas: team/FUTURE-IDEAS.txt OI-1..OI-5. Ready list: MERGE-QUEUE.txt.
2026-10-19 | ASSIGNED | Claude -> Gemini | GWRAP-300: 160+ recipes + 160+ tips for GWrap AI as gemini/gwrap-ai/knowledge-pack.json (format + real block list team/gwrap-blocks.json) - top priority
2026-10-19 | RELEASED | Claude | Beta v15.49.0 (add-on) GWRAP AI KNOWLEDGE PACKS: GWrap AI now loads recipes + tips from plain JSON packs (claude/gwrappack4900.js; format in team/GEMINI-GWRAP-AI-300-2026-10-19.txt), builds nested conditions, if/else, reporters inside inputs, variables/lists it needs; questions also match each entry's example questions. Starter pack claude/gwrap-pack.json: 8 recipes (walk+animate with arrows, gravity jump, countdown, rain clones, quiz, two-sprite talk, clicker, patrol) + 8 tips (glide, tiles, folders, forever vs repeat, broadcast, my sprite doesn't move, clone limit, song styles). claude/tests/test_gwrap_pack.js: ALL PASS. Gemini's 320+ pack merges on top.
2026-10-19 | DONE | Claude (QA agent) | Click + right-click review vs Scratch 3 / TurboWarp / PenguinMod -> claude/REVIEW-CLICKS-2026-10-19.txt: WORKS 75, BROKEN 16, MISSING 31, DIFFERENT 10. Top: new stacks have no owner sprite (move to another sprite), stage monitors hidden under the stage, undo after adding a block, clicking a script doesn't run it, reporter right-click acts on the outer block, no New message, no rename/delete variable, no Edit for custom blocks. Claude fixes these next (top of TODO).

2026-10-19 | DONE | ChatGPT | N-C1..N-C4 documentation pass for Beta v15.49.0: updated costume/animation/sounds guides, changelog and homepage highlights; app.html untouched; static GitHub edits only, no browser/live-site verification; exact 11-style list remains unverified | docs/costumes.html; docs/animation.html; docs/sounds.html; docs/changelog.html; index.html; chatgpt/AI-CHAT-LOG.txt

2026-10-19 | DONE | ChatGPT | N-C2 follow-up: verified all 11 Song maker selector labels from Claude's v15.49.0 app.html commit patch and completed docs/sounds.html; app.html untouched; GitHub/static verification only, no browser or published-site test; no review-ledger pass | docs/sounds.html; chatgpt/AI-CHAT-LOG.txt
2026-10-19 | ASSIGNED | Claude -> Gemini | CLICKS G-C1..G-C12 (new message, rename/delete variable, monitors, value bubble, edit custom block, block comments, reorder, sound menu, .sprite3, searchable dropdowns, blocks as image, copy/paste) as modules in gemini/clicks/ - then GWRAP-300
2026-10-19 | CLAIMED | Claude | engine bugs from REVIEW-CLICKS: block jumps to another sprite, monitors hidden, undo after adding a block, Ctrl+Shift+Z, reporter right-click, click a script to run + glow, sprite clicked while stopped, drag sprites on stage, Delete key
2026-10-19 | STARTED | Claude (QA agent) | full gap analysis vs Scratch, TurboWarp, PenguinMod (UI, blocks, runtime, files) -> claude/GAP-ALL-3-APPS-2026-10-19.txt
2026-10-19 | RELEASED | Claude | Beta v15.49.1 FIX: a block you just dropped no longer jumps to the next sprite you click / add / duplicate; Delete N Blocks counts right (new stacks get their sprite at once + safety net for every other way a stack is made). QA testOwner: OWN.switch/add/dup/count all WORKS.
2026-10-19 | LIST | Claude | team/MISSING-LIST.txt: everything we know G-Warp is missing or doing wrong vs Scratch 3 / TurboWarp / PenguinMod (10 broken, 12 kid-first missing, 14 TurboWarp/PenguinMod helpers), with owners; 3-app audit stopped early by the owner, its partial findings go in section E.
2026-10-19 | DONE | Claude (audit agent) | 3-app gap audit (partial, stopped by owner) -> claude/GAP-ANALYSIS-2026-10-19.txt; findings added to team/MISSING-LIST.txt section E (9 P1, 9 P2, 4 P3 groups, quick wins, NOT CHECKED list).