idea-discovery

name

idea-discovery

description

Prompt-only by default in the lightweight pack; Claude/Gemini reviewer transport is explicit opt-in. Workflow 1: Full idea discovery pipeline. Orchestrates research-lit → idea-creator → novelty-check → research-review to go from a broad research direction to validated, pilot-tested ideas. Use when user says \"找idea全流程\", \"idea discovery pipeline\", \"从零开始找方向\", or wants the complete idea exploration workflow.

Override for Codex users who want Gemini, not a second Codex agent, to act as the reviewer. Install this package after skills/skills-codex/*.

Customized Pack Defaults

This reviewer overlay is opt-in transport only. In the customized lightweight pack, default to prompt-only review unless the user explicitly asks to use this overlay reviewer backend:

• Write the review prompt under review-prompts/ and ask the user to run it in a separate conversation with the relevant skills enabled.
• Do not call Claude/Gemini reviewer MCP bridges, APIs, or follow-up status tools by default.
• Use pasted review text as the external verdict, then write compact summaries such as REVIEW_SUMMARY.md, NEXT_ACTIONS.md, or the skill-specific log.
• Keep heavy experiments AutoDL/HPC-gated and user-approved; local work stays at edits, tests, lint, dry-runs, and tiny smoke checks.

Workflow 1: Idea Discovery Pipeline

Orchestrate a complete idea discovery workflow for: $ARGUMENTS

Overview

This skill chains sub-skills into a single automated pipeline:

/research-lit → /idea-creator → /novelty-check → /research-review → /research-refine → /experiment-plan
  (survey)      (brainstorm)    (verify novel)    (critical feedback)  (refine method + plan experiments)

Each phase builds on the previous one's output. The final deliverables are a validated idea-stage/IDEA_REPORT.md with ranked ideas, plus a refined proposal (refine-logs/FINAL_PROPOSAL.md) and experiment plan (refine-logs/EXPERIMENT_PLAN.md) for the top idea.

Constants

• PILOT_MAX_HOURS = 2 — Skip any pilot experiment estimated to take > 2 hours per GPU. Flag as "needs manual pilot" in the report.
• PILOT_TIMEOUT_HOURS = 3 — Hard timeout: kill any running pilot that exceeds 3 hours. Collect partial results if available.
• MAX_PILOT_IDEAS = 3 — Run pilots for at most 3 top ideas in parallel. Additional ideas are validated on paper only.
• MAX_TOTAL_GPU_HOURS = 8 — Total GPU budget across all pilots. If exceeded, skip remaining pilots and note in report.
• AUTO_PROCEED = true — If user doesn't respond at a checkpoint, automatically proceed with the best option after presenting results. Set to false to always wait for explicit user confirmation.
• OUTPUT_DIR = idea-stage/ — All idea-stage outputs go here. Create the directory if it doesn't exist.
• REVIEWER_MODEL = gemini-review — Gemini reviewer invoked through the local gemini-review MCP bridge. Passed to the reviewer-aware sub-skills installed by this overlay.
• ARXIV_DOWNLOAD = false — When true, /research-lit downloads the top relevant arXiv PDFs during Phase 1. When false (default), only fetches metadata. Passed through to /research-lit.
• REFERENCE_LED = false — Enable evidence-grounded topic discovery from a core reference paper and optional comparators. Automatically enabled when — ref papers: is supplied.
• REF_PAPERS = false — Semicolon-separated core paper and up to three comparators. Sources accept local paths, arXiv identifiers/URLs, or paper URLs.

💡 These are defaults. Override by telling the skill, e.g., /idea-discovery "topic" — pilot budget: 4h per idea, 20h total or /idea-discovery "topic" — arxiv download: true.

Pipeline

Phase 0.5: Reference-Led Deconstruction (when REFERENCE_LED is enabled)

Skip entirely unless REFERENCE_LED = true or $ARGUMENTS contains — ref papers:.

Read and follow ../../skills-codex-library/idea-method/reference-paper-deconstruction/SKILL.md. Write idea-stage/REFERENCE_DECONSTRUCTION.md and idea-stage/TOPIC_CANDIDATES.md, then carry the candidate evidence, minimum discriminating test, and conclusion boundary into the remaining phases. Treat all statuses as provisional until /novelty-check completes.

Phase 1: Literature Survey

Invoke /research-lit to map the research landscape:

/research-lit "$ARGUMENTS"

When idea-stage/TOPIC_CANDIDATES.md exists, use it as a hypothesis ledger, not as a novelty verdict. Search broader recent literature for every carried candidate, then preserve its ID, source locators, gate status, minimum discriminating test, and conclusion boundary in the landscape notes. Report whether the lead remains open, contested, addressed, or evidence-incomplete; only /novelty-check can make the final differentiation judgment.

What this does:

• Search arXiv, Google Scholar, Semantic Scholar for recent papers
• Build a landscape map: sub-directions, approaches, open problems
• Identify structural gaps and recurring limitations
• Output a literature summary (saved to working notes)

🚦 Checkpoint: Present the landscape summary to the user. Ask:

📚 Literature survey complete. Here's what I found:
- [key findings, gaps, open problems]

Does this match your understanding? Should I adjust the scope before generating ideas?
(If no response, I'll proceed with the top-ranked direction.)

• User approves (or no response + AUTO_PROCEED=true) → proceed to Phase 2 with best direction.
• User requests changes (e.g., "focus more on X", "ignore Y", "too broad") → refine the search with updated queries, re-run /research-lit with adjusted scope, and present again. Repeat until the user is satisfied.

Phase 2: Idea Generation + Filtering + Pilots

Invoke /idea-creator with the landscape context:

/idea-creator "$ARGUMENTS"

What this does:

• If idea-stage/TOPIC_CANDIDATES.md exists, pass each retained candidate's ID, source locators, gate status, minimum discriminating evidence, and conclusion boundary into the Gemini-backed generator. These are evidence-backed seeds, not a replacement for broader candidate generation or novelty verification.
• Keep needs_evidence as an explicit search or validation requirement; keep blocked candidates out of ranking unless a narrower, evidence-backed repair is stated.
• Brainstorm 8-12 concrete ideas via the Gemini-backed /idea-creator overlay
• Filter by feasibility, compute cost, quick novelty search
• Deep validate top ideas (full novelty check + devil's advocate)
• Run parallel pilot experiments on available GPUs (top 2-3 ideas)
• Rank by empirical signal
• Output idea-stage/IDEA_REPORT.md

🚦 Checkpoint: Present idea-stage/IDEA_REPORT.md ranked ideas to the user. Ask:

💡 Generated X ideas, filtered to Y, piloted Z. Top results:

1. [Idea 1] — Pilot: POSITIVE (+X%)
2. [Idea 2] — Pilot: WEAK POSITIVE (+Y%)
3. [Idea 3] — Pilot: NEGATIVE, eliminated

Which ideas should I validate further? Or should I regenerate with different constraints?
(If no response, I'll proceed with the top-ranked ideas.)

• User picks ideas (or no response + AUTO_PROCEED=true) → proceed to Phase 3 with top-ranked ideas.
• User unhappy with all ideas → collect feedback ("what's missing?", "what direction do you prefer?"), update the prompt with user's constraints, and re-run Phase 2 (idea generation). Repeat until the user selects at least 1 idea.
• User wants to adjust scope → go back to Phase 1 with refined direction.

Phase 3: Deep Novelty Verification

For each top idea (positive pilot signal), run a thorough novelty check:

/novelty-check "[top idea 1 description]"
/novelty-check "[top idea 2 description]"

What this does:

• Multi-source literature search (arXiv, Scholar, Semantic Scholar)
• Cross-verify with the Gemini-backed /novelty-check overlay
• Check for concurrent work (last 3-6 months)
• Identify closest existing work and differentiation points

Update idea-stage/IDEA_REPORT.md with deep novelty results. Eliminate any idea that turns out to be already published.

Phase 4: External Critical Review

For the surviving top idea(s), get brutal feedback:

/research-review "[top idea with hypothesis + pilot results]"

What this does:

• Gemini acts as a senior reviewer (NeurIPS/ICML level) via the local gemini-review MCP bridge
• Scores the idea, identifies weaknesses, suggests minimum viable improvements
• Provides concrete feedback on experimental design

Update idea-stage/IDEA_REPORT.md with reviewer feedback and revised plan.

Phase 4.5: Method Refinement + Experiment Planning

After review, refine the top idea into a concrete proposal and plan experiments:

/research-refine "[top idea description + pilot results + reviewer feedback]"
/experiment-plan "[refine-logs/FINAL_PROPOSAL.md]"

What this does:

• Freeze a Problem Anchor to prevent scope drift
• Iteratively refine the method via Gemini review (up to 5 rounds, until score ≥ 9)
• Generate a claim-driven experiment roadmap with ablations, budgets, and run order
• Output: refine-logs/FINAL_PROPOSAL.md, refine-logs/EXPERIMENT_PLAN.md, refine-logs/EXPERIMENT_TRACKER.md

🚦 Checkpoint: Present the refined proposal summary:

🔬 Method refined and experiment plan ready:
- Problem anchor: [anchored problem]
- Method thesis: [one sentence]
- Dominant contribution: [what's new]
- Must-run experiments: [N blocks]
- First 3 runs to launch: [list]

Proceed to implementation? Or adjust the proposal?

• User approves (or AUTO_PROCEED=true) → proceed to Final Report.
• User requests changes → pass feedback to /research-refine for another round.
• Lite mode: If reviewer score < 6 or pilot was weak, run /research-refine only (skip /experiment-plan) and note remaining risks in the report.

Phase 5: Final Report

Finalize idea-stage/IDEA_REPORT.md with all accumulated information:

# Idea Discovery Report

**Direction**: $ARGUMENTS
**Date**: [today]
**Pipeline**: research-lit → idea-creator → novelty-check → research-review → research-refine → experiment-plan

## Executive Summary
[2-3 sentences: best idea, key evidence, recommended next step]

## Literature Landscape
[from Phase 1]

## Ranked Ideas
[from Phase 2, updated with Phase 3-4 results]

### 🏆 Idea 1: [title] — RECOMMENDED
- Pilot: POSITIVE (+X%)
- Novelty: CONFIRMED (closest: [paper], differentiation: [what's different])
- Reviewer score: X/10
- Next step: implement full experiment → /auto-review-loop

### Idea 2: [title] — BACKUP
...

## Eliminated Ideas
[ideas killed at each phase, with reasons]

## Refined Proposal
- Proposal: `refine-logs/FINAL_PROPOSAL.md`
- Experiment plan: `refine-logs/EXPERIMENT_PLAN.md`
- Tracker: `refine-logs/EXPERIMENT_TRACKER.md`

## Next Steps
- [ ] /run-experiment to deploy experiments from the plan
- [ ] /auto-review-loop to iterate until submission-ready
- [ ] Or invoke /research-pipeline for the complete end-to-end flow

Output Protocols

Follow these shared protocols for all output files:

• Output Versioning Protocol — write timestamped file first, then copy to fixed name
• Output Manifest Protocol — log every output to docs/project/OUTPUT_MANIFEST.md
• Output Language Protocol — respect the project's language setting

Key Rules

• Large file handling: If the Write tool fails due to file size, immediately retry using Bash (cat << 'EOF' > file) to write in chunks. Do NOT ask the user for permission — just do it silently.

• Don't skip phases. Each phase filters and validates — skipping leads to wasted effort later.

• Checkpoint between phases. Briefly summarize what was found before moving on.

• Kill ideas early. It's better to kill 10 bad ideas in Phase 3 than to implement one and fail.

• Empirical signal > theoretical appeal. An idea with a positive pilot outranks a "sounds great" idea without evidence.

• Document everything. Dead ends are just as valuable as successes for future reference.

• Be honest with the reviewer. Include negative results and failed pilots in the review prompt.

• Feishu notifications are optional. If ~/.codex/feishu.json exists, send checkpoint at each phase transition and pipeline_done at final report. If absent/off, skip silently.

Composing with Workflow 2

After this pipeline produces a validated top idea:

/idea-discovery "direction"         ← you are here (Workflow 1, includes method refinement + experiment planning)
/run-experiment                     ← deploy experiments from the plan
/auto-review-loop "top idea"        ← Workflow 2: iterate until submission-ready

Or use /research-pipeline for the full end-to-end flow.