idea-discovery

name

idea-discovery

description

Workflow 1: Full idea discovery pipeline. Orchestrates research-lit → idea-creator → novelty-check → research-review → research-refine plus experiment-plan to go from a broad research direction to validated, pilot-tested ideas. Also handles robotics / embodied AI directions through simulation-first robotics mode. Use when user says \"找idea全流程\", \"idea discovery pipeline\", \"从零开始找方向\", \"robotics idea discovery\", \"embodied AI idea\", or wants the complete idea exploration workflow.

Workflow 1: Idea Discovery Pipeline

Capability Routing

This is a first-layer entry skill. Keep it loaded as the user-facing route; when a request needs a specialized capability below, resolve the debuffer repo root from .debuffer_skills/installed-skills-codex.txt (repo_root) when available, read the referenced library SKILL.md, then follow that skill. Do not copy the whole library skill into this file.

• /idea-creator: read ../../skills-codex-library/idea-method/idea-creator/SKILL.md.
• /reference-paper-deconstruction: read ../../skills-codex-library/idea-method/reference-paper-deconstruction/SKILL.md.
• /research-refine: read ../../skills-codex-library/idea-method/research-refine/SKILL.md.
• /research-lit: read ../../skills-codex-library/literature/research-lit/SKILL.md.
• /novelty-check: read ../../skills-codex-library/review/novelty-check/SKILL.md.

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)  (method)        (experiment plan)

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.

Customized Pack Defaults

Read ../shared-references/lightweight-research-pack.md first. Default to lightweight idea discovery unless the user requests a full report:

• Classify startup mode (venue-only, reference-paper, reference-codebase, idea-doc, existing-repo, or partial-results) and tailor the first deliverable to the available material.
• AUTO_PROCEED = false, COMPACT = true, RENDER_HTML = false.
• For venue-only, create a concise PROJECT_BRIEF.md, idea-stage/IDEA_CANDIDATES.md, and NEXT_ACTIONS.md before any long report.
• For paper/code/reference starts, extract assumptions, reusable assets, and the smallest validation plan instead of generating a large IDEA_REPORT.md by default.
• Do not run heavy pilots locally. If a pilot needs GPU time, prepare an AutoDL-ready smoke/formal plan and ask for approval.
• External critique is prompt-only by default: write review-prompts/idea_review_prompt.md and wait for pasted feedback.
• Treat idea-creator as the owner of A-E candidate generation. This workflow should only pass startup mode, landscape, reference material, venue constraints, and failure-memory context into idea-creator.
• When REFERENCE_LED = true, run reference-paper-deconstruction before the broad landscape search. It produces evidence-backed candidate seeds; it does not replace idea-creator, novelty-check, or the user checkpoint.
• Treat every reference-led candidate as provisional until its novelty check passes; do not turn a future-work statement into a confirmed gap.
• If the direction is robotics or embodied AI, use robotics mode: simulation first, benchmark-specific ideas only, explicit embodiment/task/action frame, failure metrics, and no real-robot execution without user approval.

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.
• REVIEWER_MODEL = gpt-5.5 — Model used via a secondary Codex agent. Must be an OpenAI model (e.g., gpt-5.5, o3, gpt-4o). Passed to sub-skills.
• 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.
• COMPACT = false — When true, generate compact summary files for short-context sessions and downstream skills. Writes idea-stage/IDEA_CANDIDATES.md.
• OUTPUT_DIR = idea-stage/ — All idea-stage outputs go here. Create the directory if it doesn't exist.
• REF_PAPER = false — Legacy single-paper fast path. Accepts a local PDF path, arXiv URL, or paper URL and writes idea-stage/REF_PAPER_SUMMARY.md.
• 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.
• RENDER_HTML = true — When true (default), auto-render idea-stage/IDEA_REPORT.md to HTML at workflow end via /render-html. Uses --no-review (source already passed novelty + cross-model review during Phase 3). Set false to skip, or pass — render html: false.

💡 These are defaults. Override by telling the skill, e.g., /idea-discovery "topic" — ref paper: https://arxiv.org/abs/2406.04329 or /idea-discovery "topic" — compact: true.

Pipeline

Phase 0: Load Research Brief (if available)

Before starting any other phase, check for a detailed research brief in the project:

1. Look for docs/project/RESEARCH_BRIEF.md or a path passed in $ARGUMENTS. Fall back to legacy root RESEARCH_BRIEF.md only for older projects.
2. If found, read it and extract:
   • problem statement and context
   • constraints: compute, data, timeline, venue
   • what the user already tried and what did not work
   • domain knowledge and non-goals
   • existing results, if any
3. Use this as the primary context for all subsequent phases; it replaces the one-line prompt when more specific.
4. If both docs/project/RESEARCH_BRIEF.md and one-line $ARGUMENTS exist, merge them: the brief has priority for details, and the argument sets the direction.

If no brief exists, proceed normally with $ARGUMENTS as the research direction.

Recommended template:

# Research Brief

## Problem Statement
[What problem are we trying to solve?]

## Context
[Relevant field, current approach, why this matters]

## Constraints
- Compute:
- Data:
- Timeline:
- Target venue:

## What We Already Tried
- [attempt] -> [outcome]

## Non-Goals
- [what not to pursue]

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

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

1. Invoke /reference-paper-deconstruction with the research direction and the supplied reference set. A single core paper is sufficient; the skill may add comparators selected for question/design/data compatibility.
2. Write and read idea-stage/REFERENCE_DECONSTRUCTION.md and the compact idea-stage/TOPIC_CANDIDATES.md handoff.
3. Carry the candidate evidence, minimum discriminating test, and conclusion boundary into Phases 1 and 2. A ready_for_novelty_check status is not a novelty verdict; needs_evidence and blocked records remain constraints, not ideas to silently promote.
4. Do not generate a legacy REF_PAPER_SUMMARY.md unless the user requested REF_PAPER without reference-led mode. This preserves the old fast path.

Phase 0.6: Reference Paper Summary (when REF_PAPER is set)

Skip entirely if REF_PAPER is false.

Summarize the reference paper before searching the literature:

1. If arXiv URL — invoke /arxiv "ARXIV_ID" — download to fetch the PDF, then read the first 5 pages.
2. If local PDF path — read the PDF directly, focusing on the title, abstract, introduction, and method overview.
3. If other URL — fetch the content and extract the method, results, and limitations.
4. Generate idea-stage/REF_PAPER_SUMMARY.md using this template:

# Reference Paper Summary

## What They Did
[2-3 sentences: core method and contribution]

## Key Results
[Main quantitative findings]

## Limitations & Open Questions
[Acknowledged weaknesses, missing experiments, future work]

## Potential Improvement Directions
[Concrete ways to extend, challenge, or improve the paper]

## Codebase
[If `base repo` is set: link to the repo and identify relevant entry points]

Use idea-stage/REF_PAPER_SUMMARY.md as additional context in both Phase 1 and Phase 2.

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. For every carried candidate, search broader recent literature to determine whether its stated gap is still open, contested, already addressed, or still missing evidence. Preserve the candidate ID, source locators, gate status, minimum discriminating test, and conclusion boundary in the landscape notes. This check may downgrade a lead, but 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 and any available idea-stage/REF_PAPER_SUMMARY.md or idea-stage/TOPIC_CANDIDATES.md handoff:

/idea-creator "$ARGUMENTS"

What this does:

• If idea-stage/REF_PAPER_SUMMARY.md exists, include it as context so ideas explicitly build on, improve, or extend the reference paper
• If idea-stage/TOPIC_CANDIDATES.md exists, pass each retained candidate's ID, source locators, gate status, minimum discriminating evidence, and conclusion boundary. These are evidence-backed seeds, not a replacement for A-E generation or novelty verification.
• Keep needs_evidence as an explicit search or validation requirement; do not turn it into an accepted premise. Keep blocked candidates out of ranking unless a narrower, evidence-backed repair is stated.
• Delegate A-E candidate generation and compact idea-memory updates to idea-creator; do not restate that logic here.
• Pass idea-stage/IDEA_MEMORY.md and experiments/NEGATIVE_RESULTS.md as banlist context unless the user explicitly asks to revisit a failed direction.
• Brainstorm 8-12 concrete ideas via GPT-5.4 xhigh
• Filter by feasibility, compute cost, quick novelty search
• Deep validate top ideas (full novelty check + devil's advocate)
• Prepare pilot specs for top ideas. Do not run heavy pilots locally; when GPU time is needed, create AutoDL-ready pilot command blocks and wait for approval or pasted results.
• 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 GPT-5.4 xhigh
• 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:

• GPT-5.4 xhigh acts as a senior reviewer (NeurIPS/ICML level)
• 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:

Run /research-refine first, then /experiment-plan when the method thesis is stable:

/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 GPT-5.4 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

Phase 5.5: Write Compact Files (when COMPACT = true)

Skip entirely if COMPACT is false.

Write idea-stage/IDEA_CANDIDATES.md — a lean summary of the top 3-5 surviving ideas:

# Idea Candidates

| # | Idea | Pilot Signal | Novelty | Reviewer Score | Status |
|---|------|-------------|---------|---------------|--------|
| 1 | [title] | +X% | Confirmed | X/10 | RECOMMENDED |
| 2 | [title] | +Y% | Confirmed | X/10 | BACKUP |
| 3 | [title] | Negative | — | — | ELIMINATED |

## Active Idea: #1 — [title]
- Hypothesis: [one sentence]
- Key evidence: [pilot result]
- Next step: /experiment-bridge or /research-refine

Update idea-stage/IDEA_MEMORY.md append-only for every accepted, rejected, pilot-pass, pilot-fail, or inconclusive idea using the template in autosci-lite-patterns.md. Keep the entry short and specific enough that future sessions know what not to repeat.

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

Render HTML view (auto, when RENDER_HTML = true)

After finalizing idea-stage/IDEA_REPORT.md (and the optional IDEA_CANDIDATES.md), invoke /render-html on the report so the user has a single-file HTML view for tablet / phone reading:

/render-html "idea-stage/IDEA_REPORT.md" --no-review

--no-review is intentional: source MD already passed this skill's own novelty + cross-model review. HTML render is a structural conversion, not a new claim-audit gate.

Non-blocking: if /render-html fails (helper missing, secondary Codex agent unavailable, file write error), log the failure and continue. Skip entirely if RENDER_HTML = false.

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.