wt-idea

WT Idea

Use this skill to capture and enrich ideas before they are ready for wt-ready. The goal is to preserve enough evidence and context that a later wt-ready pass can turn the idea into the best available task/workflow plan.

Do not implement code, create TaskDocuments, create workflows, launch worktrees, or decide final scope from this skill. Use wt-ready for idea-to-task conversion, wt-start for execution, wt-coordinate for running work, and wt-land for reviewed work.

First Read

Inspect local truth before asking questions:

git status --short --branch
find . -maxdepth 2 -name AGENTS.md -o -name AGENTS.override.md
common_dir="$(git rev-parse --git-common-dir)"
find "$common_dir/wt/ideas" "$common_dir/wt/tasks" "$common_dir/wt/workflows" -maxdepth 1 -type f 2>/dev/null | sort
wt config 2>/dev/null || true

Never read secret files such as .env. If <git-common-dir>/wt/ideas does not exist, create it only when saving a new idea.

Capture Standard

An idea is not a task. Capture it as a durable research artifact with:

• the user's raw intent, preserving wording when useful
• the user/customer problem or desired outcome
• relevant local product/code/docs context
• related existing ideas, tasks, workflows, issues, or docs
• external best-practice references when the decision depends on current or domain-specific practice
• possible solution directions, with tradeoffs
• assumptions that need validation
• risks, rabbit holes, and explicit non-goals
• open questions for a later HITL or wt-ready pass
• a recommendation for the next step: enrich more, run wt-ready, defer, or archive

Prefer information density over premature certainty. A good wt-idea output makes later decisions better; it does not force one solution too early.

Evidence Gathering

Start from the conversation and repository. Search existing local artifacts before creating a new one:

common_dir="$(git rev-parse --git-common-dir)"
rg -n "<keyword>|<related term>" "$common_dir/wt/ideas" "$common_dir/wt/tasks" docs app resources tests 2>/dev/null

Use external research when the user asks for best practices, the idea concerns current tooling/frameworks, or the best direction cannot be judged from the repo alone. Prefer primary or authoritative sources: official docs, established method writeups, source repositories, standards, and vendor docs. Record URLs in the idea body.

Useful discovery lenses:

• Product discovery: outcome -> opportunity -> solution -> experiment.
• Feedback synthesis: raw feedback -> insight -> related idea.
• Triage inbox: review, dedupe, label, clarify, then accept/defer/archive.
• Shape Up style shaping: problem, appetite, solution sketch, rabbit holes, no-gos.
• Technical decision records: context, options, decision drivers, consequences.

Separate confirmed facts, source-backed guidance, inference, and unresolved questions.

Status Model

Use these statuses in idea files:

• captured: raw idea saved with minimal context.
• enriched: local/external context and alternatives have been gathered.
• ready_for_wt_ready: enough information exists for wt-ready to prepare TaskDocuments/workflows.
• converted: already turned into TaskDocuments or workflow by wt-ready.
• archived: intentionally not pursuing now.

Default to enriched when you performed meaningful research. Use ready_for_wt_ready only when scope, unresolved questions, and next checks are clear enough for wt-ready to proceed without rediscovering the basics.

File Format

Store ideas in <git-common-dir>/wt/ideas/<date>-<slug>.toml. Use lowercase ASCII kebab-case slugs. If an idea already exists, update the existing file instead of creating a duplicate.

Use only simple top-level TOML fields and put rich planning context in body:

title = "Short Korean title"
status = "enriched"
created_at = "YYYY-MM-DD"
updated_at = "YYYY-MM-DD"
source = "user"
tags = ["wiki", "product-discovery"]

body = """
Raw intent:
- ...

Outcome / problem:
- ...

Evidence:
- Local: ...
- External: ...

Options:
- Option A: ...
- Option B: ...

Tradeoffs:
- ...

Risks / rabbit holes:
- ...

Non-goals:
- ...

Open questions:
- ...

Next step:
- Run wt-ready on this idea when ...
"""

Do not add nested TOML tables unless the local schema explicitly adopts them. Keep the file easy to read, diff, and convert later.

Questions

Ask only when the idea cannot be captured safely without an answer. Otherwise, record uncertainty as an open question and proceed.

Ask at most one focused question at a time. Include the recommended default.

Handoff

End with:

• idea file path
• status
• evidence checked
• duplicates or related artifacts found
• why it is or is not ready for wt-ready
• exact next skill invocation or target, for example:

$wt-ready <git-common-dir>/wt/ideas/YYYY-MM-DD-slug.toml

If the user asked only for a list or review of ideas, do not write files unless they explicitly ask to register or update an idea.