| name | webgpt-todo-response |
| description | Use after receiving WebGPT or another LLM review on a Formax todo and before sending the todo back for another pass. Produce a concise handoff response that says what we adopted, what we reject or question, and what the reviewer should specifically re-evaluate in `docs/todolist.md`. |
WebGPT Todo Response
Purpose
Generate the message to send back to WebGPT after we have read its previous
analysis and drafted or updated docs/todolist.md.
Use this skill when the user asks:
- whether we have rebuttals or questions for WebGPT
- what to include when sending our todo back to WebGPT for another pass
- to prepare a response asking WebGPT to evaluate, improve, or challenge our todo
- to compare WebGPT's recommendations against our chosen implementation scope
Inputs To Inspect
Read only the files needed for the current handoff:
- WebGPT response, usually under
repomix-output/
- current todo, usually
docs/todolist.md
- relevant canonical docs under
docs/contracts/*, docs/frontend/*, or other explicitly governing docs when the todo depends on them
- optional other LLM replies if the user asks for a multi-model synthesis
Do not re-run broad repository analysis unless the todo or WebGPT response
depends on code facts that are unclear.
Workflow
-
Identify WebGPT's strongest recommendations.
- Mark which ones are adopted in the todo.
- Mark which ones are intentionally deferred.
- Mark which ones are rejected or still need clarification.
-
Check the todo against Formax boundaries.
- Canonical semantics belong in
docs/contracts/* and canonical runtime layers, not only UI.
- Web reference UI should reflect runtime/platform truth, not invent it.
- Do not move thread/runtime state ownership into ad hoc component-local logic when the task is structurally runtime-driven.
- Preserve parity-sensitive behavior when relevant: transcript surface semantics, URL/thread sync, prompt/tool exposure boundaries, permissions flow, and active-thread canonical gating.
- Avoid turning a focused task into a broad cleanup or cross-subsystem redesign unless explicitly requested.
-
Find weak spots in the todo.
- Missing canonical-doc step
- Missing data/type/interface step before UI
- Runtime state ownership drift
- Welcome/draft/thread semantics being mixed together
- Scope creep into unrelated app-server, terminal, diff, approval, or desktop integration work
- Missing tests or review gates
- Missing statement of protocol constraints or non-atomic failure boundaries
-
Write a concise message for WebGPT.
- Assume WebGPT has no hidden context beyond the attached todo and bundle.
- Be explicit about decisions already made.
- Ask targeted questions instead of open-ended “any thoughts?”
- Request concrete todo edits or challenges, not generic feedback.
Output Shape
Produce a copy-ready Markdown response with these sections:
# Response To WebGPT
## What We Adopted
- ...
## Where We Differ / Pushback
- ...
## My Current Leaning
1. ...
## Highest-Value Review Points
1. ...
## Specific Questions For You
1. ...
## Please Review The Todo For
- ...
## Constraints To Preserve
- ...
Keep it short enough to paste into WebGPT with the todo. Prefer 5-10 specific
questions/checks over a long essay.
Use My Current Leaning to distinguish default decisions from genuinely open
questions. WebGPT may challenge these, but should not treat them as blank slate.
Use Highest-Value Review Points to focus WebGPT on the few risks most likely
to improve the todo. These should be sharper than the broader checklist.
Good Question Patterns
- “Does this todo still hide new semantics inside
!activeThreadId, or is the draft state truly first-class?”
- “Are we separating
selectedCwd from draftCwd cleanly enough to avoid left-rail/runtime state drift?”
- “Is the proposed first-send flow realistic given
thread/start and turn/start are non-atomic?”
- “Are we over-expanding the task into unrelated desktop/add-project behavior instead of keeping the mainline on new-thread draft semantics?”
- “Do the loops lock runtime ownership first, then UI, then tests, or is there still UI-first drift?”
Avoid
- Do not ask WebGPT to implement patches unless the user explicitly wants that.
- Do not ask WebGPT to run commands.
- Do not include local absolute paths.
- Do not send vague requests like “please improve this.”
- Do not restate the whole todo; reference it and ask for specific audit points.