Skip to main content
skill CX Codex

ccf-idea-reviewer — Claude Code skill for idea review

The ccf-idea-reviewer Claude Code skill assesses research ideas for value, novelty, insight, and mechanism. It reviews concepts by default, without judging experiments.

MIT 2,200 installs ★ 3,769 updated 2026-10-20

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add mikubaka88/CCFA-Skills --skill ccf-idea-reviewer -a claude-code
Open on GitHub

What ccf-idea-reviewer does

The skill examines a research idea by mapping its problem, gap, insight, and mechanism, then judging its importance, novelty, coherence, and audience fit. It uses ccf-literature-searcher to resolve gaps in evidence about nearby work and checks primary sources before making overlap claims. If browsing is forbidden, the review records the search coverage accordingly.

The default scope is conceptual. Experiments, baselines, and implementation resources do not affect the verdict unless that assessment is requested separately. Each major criticism points to a specific claim or assumption and gives the smallest repair; uncertain novelty is kept distinct from confirmed overlap.

A standard request produces a detailed report with a verdict, prior-work comparison, anchored concerns, a scorecard, and confidence. Idea development belongs to ccf-idea-optimizer; manuscript evidence and writing are handled by ccf-paper-reviewer. Before file work, the skill requires applying ccf-humanization and ccf-common.

What it helps with

  • judge a research idea against nearby work
  • trace a gap between the problem, insight, and mechanism
  • compare several concepts under one set of criteria

Example prompts for the agent

  • Assess the novelty and importance of this research idea.
  • Compare these concepts and explain which mechanism is more coherent.
  • Review the central idea in this manuscript without assessing its experiments.

How to install ccf-idea-reviewer

1Skills CLI (Vercel)

npx skills add mikubaka88/CCFA-Skills --skill ccf-idea-reviewer -a claude-code

2Manually: put the files in place

~/.claude/skills/ccf-idea-reviewer/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/ccf-idea-reviewer/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about ccf-idea-reviewer

How do I install ccf-idea-reviewer in Claude Code?

Run npx skills add mikubaka88/CCFA-Skills --skill ccf-idea-reviewer -a claude-code. The skill file is at ~/.claude/skills/ccf-idea-reviewer/SKILL.md.

What does ccf-idea-reviewer do?

It assesses research concepts for importance, novelty, insight, and mechanism coherence. Experiment results and implementation resources are outside its default scope.

How does ccf-idea-reviewer differ from ccf-paper-reviewer?

This skill reviews the idea itself. ccf-paper-reviewer assesses manuscript evidence, scientific completeness, and writing; ccf-idea-optimizer develops ideas.

Source text

SKILL.md · MIT
---
name: ccf-idea-reviewer
description: "Assess research ideas for value, novelty, insight, and mechanism. Use for 思路审核, 靠谱吗, 值得做吗, 创新够不够, idea review, scoring, and ranking; no numeric-score request is needed. Default to concept-only review without experiment assessment, including ideas extracted from manuscripts. Developing an idea belongs to ccf-idea-optimizer; evaluating manuscript evidence or writing belongs to ccf-paper-reviewer."
metadata:
  ccf_skill_controls:
    handoff_question_mode: partial
    respect_session_denylists: true
    protect_idea_scope_in_writing: true
    private_material_safety: moderate
    shared_controls: ../ccf-common/references/
---

# CCF Idea Reviewer

## Family File Contract

Before file work, reuse canonical paths and one task/artifact working directory; default to project-root `ccfa-workfiles/<purpose>/<artifact-id>/`. Update in place, preserve inputs, and use UTF-8 with Chinese-text checks. Apply [artifact-contracts.md](../ccf-common/references/artifact-contracts.md) for naming, ownership, and task-created cleanup; handoffs inherit these paths.

## Collaboration Contract

Before specialist execution, read and apply [ccf-humanization](../ccf-humanization/SKILL.md) first, then [ccf-common](../ccf-common/SKILL.md). At every handoff, reuse their applicable active rules or refresh missing/changed ones. Both preflights are required even without prose; detailed editing, experiment, and maintenance modes run only when relevant.

Integrate necessary specialist contributions under one owner; reuse valid prerequisites and verify affected results. Follow the [cooperation routes](../ccf-common/references/routing.md) without duplicate reports.

## Invocation Controls

**CCFA Handoff Mode: PARTIAL (Recommended).** Follow `metadata.ccf_skill_controls.handoff_question_mode`, `../ccf-common/references/handoff-modes.md`, and `../ccf-common/references/task-modes.md`. Infer assessment from the requested judgment; an exact skill name, review keyword, or numeric score is unnecessary. A rough idea can still receive a serious concept assessment.

Choose by the object of judgment, not file type: a PDF supplied for “只看核心思路” remains idea review. Use `ccf-idea-optimizer` for requested development and `ccf-paper-reviewer` for manuscript evidence, scientific completeness, writing, or version readiness. Execute an explicitly combined review/development task through its respective owners without another permission round.

## Core Rule

Judge problem importance, novelty against closest work, conceptual insight, mechanism coherence, elegance, and audience fit. Default to concept-only scope. Do not grade experiments, demand baselines/ablations/results, assess implementation resources, or lower the verdict because research is unfinished. Add experiment or feasibility assessment only when the user requests that extension, keeping it separate from the concept score.

Distinguish a logical contradiction from an untested hypothesis. A meaningful claim can be assessed before experimental validation. Separate concept quality, development potential, and confidence; do not turn uncertain novelty into demonstrated overlap or submission readiness into idea quality.

Every consequential criticism identifies the affected idea statement or assumption, its inspected basis, its significance, and the smallest repair. Do not fabricate prior art, results, reviewer agreement, or acceptance probability. Use `abandon` only after identifying why no meaningful formulation or plausible rescue remains.

The smallest repair explains a diagnosis; it does not limit an authorized optimizer to incremental changes. Review changed assumptions and unresolved concerns; unchanged evidence does not require another full scorecard. Converge on resolved findings or the exact remaining discriminator, not repeated scoring until a preferred number appears.

## Workflow

For an internal contribution, run the checks needed for its assigned conceptual question and evidence dependencies, then return findings under the internal output contract. The complete workflow and report apply to a user-requested idea review.

1. Identify the requested judgment, concept, audience, available sources, and any explicit scope extension. Reuse conversation context; ask only for a missing decision that changes the assessment. Do not ask for experimental materials merely to start idea review.
2. Load `references/strict-idea-review.md` for report selection and assessment, including qualitative judgments. Normalize problem → gap → insight → mechanism; keep experimental planning outside default intake.
3. Ground decisive novelty claims through public-safe retrieval under `../ccf-common/references/privacy-and-evidence.md`, unless browsing is forbidden. Use `ccf-literature-searcher` to resolve missing closest-work evidence and integrate its mechanism comparison before the verdict. Reuse applicable verified sources and record searched, partially searched, supplied-only, or unsearched coverage. Inspect relevant primary-source content before claiming overlap; missing experimental results are not a concept-review prerequisite.
4. Assess distinct conceptual perspectives using `references/expert-panel.md`; combine duplicate issues under stable IDs. Experiment reviewers are optional for a requested extension. Use independent calls only if permitted and useful, and label single-agent perspectives honestly.
5. For standard scoring, load `references/rubric.md`, `references/calibration.md`, and `../ccf-common/references/review-output-standards.md`. Use the six conceptual dimensions and assessed-weight coverage. Honor no-score requests with qualitative judgments; low confidence is not a low score.
6. Distinguish decisive conceptual flaws from repairable gaps and unanswered questions. Compare multiple ideas under a common scope and rubric. Re-review changed assumptions and unresolved concerns without imposing new experiment criteria.
7. For a user-requested review, deliver the detailed report from `strict-idea-review.md` by default; use its brief version only for an explicit brevity request or restrictive user format. Put requested optimization or experiment work in its own authorized deliverable.

## Output Contract

For a bounded internal concept check requested by another skill, return the inspected concept, evidence-backed findings, unresolved questions, and completion conditions to its owner. Do not create a standalone score report or develop a replacement idea. A user-requested idea review retains the report requirements below, including its detailed default and concept-only scope.

Use the concept-review structure defined in `references/strict-idea-review.md`; do not substitute a generic coaching response or manuscript acceptance report. State the conceptual verdict, prior-art delta, anchored concerns, applicable scorecard, development potential, confidence, and concrete refinements without repeating the same criticism. A rough seed, short prompt, or no-score request does not select brief output or authorize experiment assessment.

For an explicitly requested brief judgment, use the template's five blocks and retain the same concept-only boundary. No forced scores or experimental checklist. Recommendations remain `accept-to-develop`, `revise`, `pivot-with-rescue-route`, `abandon`, or `needs-literature-search`.

## References

- `references/strict-idea-review.md`: standard report structure, grounding, and scope.
- `references/rubric.md`, `references/calibration.md`: concept dimensions, weights, coverage, and decision conditions.
- `references/expert-panel.md`: distinct conceptual perspectives and optional requested extensions.
- `references/source-notes.md`: public provenance and reuse boundaries.
- `../ccf-common/references/review-output-standards.md`: evidence, scoring, and concern continuity.

For file outputs, follow `../ccf-common/references/artifact-contracts.md`: reuse the established report path, keep intermediate files under one stable task directory, and update current files in place. Load this policy only when writing files and it is not already in context.

© mikubaka88 · MIT · mikubaka88/CCFA-Skills

Similar skills

All skills for Science and research →

From the same set18

The whole set →
ccf-pipeline-orchestrator The ccf-pipeline-orchestrator Claude Code skill plans research stages and coordinates specialist skills. It handles task breakdown, project tracking, and explicitly requested autonomous research. CX Codex +1 2,200installs ccf-integrity-auditor ccf-integrity-auditor is a Claude Code skill that checks claims, figures, terms, and citations against supplied evidence. It audits manuscripts and research results. CX Codex +1 2,200installs ccf-literature-monitor The ccf-literature-monitor Claude Code skill tracks recent research and checks papers for overlap with an idea. It scans sources such as arXiv, OpenReview, conference pages, labs, and named competitors. CX Codex +1 2,200installs ccf-visual-composer ccf-visual-composer renders Python data analyses and generates scientific diagrams through a three-stage GPT Image workflow. Triggered when building architecture diagrams or redrawing figures. CX Codex +1 2,200installs ccf-skill-forger ccf-skill-forger — a skill for Claude Code, Codex, Cursor and other agents that audits and maintains the CCFA skill family: structure, triggers, dependencies, and SVG diagrams. CX Codex +1 2,200installs ccf-project-scaffolder ccf-project-scaffolder scaffolds academic research directories, LaTeX templates for CCF venues, and ccfa.yaml configuration. Runs during repository setup without altering existing manuscript files. CX Codex +1 2,200installs ccf-paper-to-exemplar ccf-paper-to-exemplar turns supplied paper PDFs into writing exemplar cards: argument structure, reusable techniques, do-not-copy boundaries. Cards go into a personal library that other CCFA skills reuse. CX Codex +1 2,200installs ccf-research-engineer The ccf-research-engineer Claude Code skill supports research code selection, implementation, and training diagnosis. It enters the workflow for baseline reproduction, model changes, dependency conflicts, or memory and throughput issues. CX Codex +1 230installs

Project sponsors

Companies that keep this analytics open