idea-discovery skill

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-pipeline
  (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/IDEAREPORT.md with ranked ideas, plus a refined proposal (refine-logs/FINALPROPOSAL.md) and experiment plan (refine-logs/EXPERIMENT_PLAN.md) for the top idea.

Constants

• PILOTMAXHOURS = 2 — Skip any pilot experiment estimated to take > 2 hours per GPU. Flag as "needs manual pilot" in the report.
• PILOTTIMEOUTHOURS = 3 — Hard timeout: kill any running pilot that exceeds 3 hours. Collect partial results if available.
• MAXPILOTIDEAS = 3 — Run pilots for at most 3 top ideas in parallel. Additional ideas are validated on paper only.
• MAXTOTALGPUHOURS = 8** — Total GPU budget across all pilots. If exceeded, skip remaining pilots and note in report.
• AUTOPROCEED = true** — When true, checkpoints are informational: report the selected option and continue in the same turn. Set to false to ask for explicit user confirmation and end the turn at each selection checkpoint.
• REVIEWERMODEL = gpt-6-astra** — Model used via a secondary Codex agent. Must be an OpenAI model (e.g., gpt-6-astra, o3, gpt-4o). Passed to sub-skills.
• ARXIVDOWNLOAD = 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.
• OUTPUTDIR = idea-stage/** — All idea-stage outputs go here. Create the directory if it doesn't exist.
• REFPAPER = false** — Reference paper to base ideas on. Accepts a local PDF path, arXiv URL, or paper URL. When set, summarize it first and use it as idea-generation context.
• RENDERHTML = true — When true (default), auto-render idea-stage/IDEAREPORT.md to HTML at workflow end via /render-html. Uses --no-review because the source already received novelty + same-family provisional review. Set false to skip.
• RESUMABLE = true — Record stage evidence under .aris/runs/.json and require a deterministic evidence gate before declaring the final report complete.

💡 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.

Checkpoint execution rule

Resolve AUTO_PROCEED once from $ARGUMENTS before Phase 0 and keep that mode for the entire workflow.

not a question. State the result and the automatically selected next action, then continue executing in the same turn. Do not ask for confirmation, request user input, sleep, wait for silence, or end the turn at a checkpoint.

• AUTOPROCEED=true is non-blocking.** A checkpoint is a progress update,

end the turn. Resume only after an explicit reply.

• AUTOPROCEED=false is blocking.** Present the options, ask the user, and

Never implement auto-proceed as “ask, then continue if there is no response.” Once a turn ends, silence cannot resume the workflow. The user can still interrupt a non-blocking run at any time.

This rule governs only AUTOPROCEED-controlled selection checkpoints. If the user explicitly enables a Feishu interactive gate, that external approval or reply is an intentional blocking exception; wait for that user-controlled gate rather than treating it as a silence timeout. Feishu off/push-only modes remain non-blocking under AUTOPROCEED=true.

Per-stage evidence gate (RESUMABLE = true)

Resolve runstate.py and ideadiscovery_gate.py from the Codex manifest using the same resolver pattern as /research-pipeline. If either helper is unavailable, the final report is BLOCKED; do not silently continue without a state record.

For a new run, derive from the direction slug and date, then start this ordered state record with --executor --provisional-advances (for example, codex-gpt-6-astra):

research-lit,idea-creator,novelty-check,research-review,research-refine-pipeline

For each phase, mark running on entry and done --artifact only after its artifact is present. Use these artifact locators so the final gate can check the canonical report rather than scattered scratch files:

novelty-check and research-review are reviewer-bearing phases. A done status or a heading alone is not review evidence. After each phase has folded substantive findings into its anchored report section, first record it done, then, only after the secondary Codex reviewer actually returns a positive, identity-bearing verdict, record its honest same-family receipt using the actual reviewer model and durable agent/trace id:

python3 <resolved-run_state.py> mark-provisional . <run_id> novelty-check --verdict-id "<agent-or-trace-id>" --reviewer "<actual-Codex-reviewer-model>"
python3 <resolved-run_state.py> mark-provisional . <run_id> research-review --verdict-id "<agent-or-trace-id>" --reviewer "<actual-Codex-reviewer-model>"

Never invent either value and never mark a phase provisional without the positive verdict required by the run-state contract. For novelty-check, both PROCEED and PROCEED WITH CAUTION are positive verdicts — caution is guidance for the pilot, not a rejection; only ABANDON is negative. A negative verdict does not grant a review receipt. Leave the phase done and the final gate BLOCKED, select a surviving or new idea, then re-run that reviewer-bearing phase. Do the same if the reviewer is unavailable, returns no valid identity/response, or its output was not folded into the report. --provisional-advances is required because these same-family receipts are explicitly provisional, not cross-family acceptance.

If a reviewer overlay actually returns a recognized different-family model (for example Claude reviewing a Codex run), use accept with that overlay's real model and trace id instead. Do not mislabel a cross-family receipt as provisional; the evidence gate validates either honest route from the recorded families.

At the end of Phase 5, run:

python3 <resolved-idea_discovery_gate.py> . <run_id> --report idea-stage/IDEA_REPORT.md

The gate writes its result to gates.idea-discovery-evidence in the run state. On PASS, it has validated (but never created) the two review receipts, all required artifacts, and non-empty anchored report sections. Per-phase acceptance or provisional status stays with the reviewer route that produced the receipt. On a non-zero exit, the gate writes explicit BLOCKED: evidence missing lines to the report; do not present the workflow as complete. On — resume , start from the first non-terminal phase and re-run the gate before finalizing.

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 RESEARCH_BRIEF.md in the project root or a path passed in $ARGUMENTS.
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

1. Use this as the primary context for all subsequent phases; it replaces the one-line prompt when more specific.
2. If both 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 Paper Summary (when REF_PAPER is set)

Skip entirely if REFPAPER 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/REFPAPERSUMMARY.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/REFPAPERSUMMARY.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" — composed: idea-stage/IDEA_REPORT.md

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.

When AUTOPROCEED=true (non-blocking):** report the selected direction and continue immediately in the same turn, without a question:

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

AUTO_PROCEED: selected [top-ranked direction]. Continuing to Phase 2.

When AUTOPROCEED=false (blocking):** present the same findings, ask Does this match your understanding? Should I adjust the scope before generating ideas?, then end the turn.

• User approves → proceed to Phase 2 with the 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 idea-stage/REFPAPERSUMMARY.md if available:

/idea-creator "$ARGUMENTS" — composed: idea-stage/IDEA_REPORT.md

What this does:

• If idea-stage/REFPAPERSUMMARY.md exists, include it as context so ideas explicitly build on, improve, or extend the reference paper
• Brainstorm 8-12 concrete ideas via GPT-6-Astra xhigh
• 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.