Claude Skills by shenmuxing

This tutorial creates a minimal reusable skill package.

This tutorial converts a private local skill into a public-safe candidate.

Review MetaSkill skills before publication for privacy, public-safe examples, frontmatter and folder standards, truthful capability claims, and resource consistency. Use before moving a skill into skill-examples, publishing a skill, importing a local skill into MetaSkill, or responding to requests for public-readiness review.

Turn a theory-first research idea into a proof-oriented research plan. Use after idea-creator-analysis, or when the user asks for theorem targets, assumptions, proof obligations, impossibility routes, analysis validation, or a non-experimental plan.

Use after the user has authorized GPT Pro help; manage a prompt-plus-sources workspace, route through ChatGPT Projects when available, and audit returned reasoning.

Protocol-level two-agent paper writing workflow. Use when the user wants Codex to plan and review from global context while DeepSeek writes or revises the manuscript through the global deepseek-agent skill, with muxing-style-review used for explicit prose style checks.

Call a DeepSeek-backed OpenCode agent as a separate critique, writing, or revision agent from Codex. Use when Codex needs to delegate adversarial research critique, novelty skepticism, method review, manuscript prose, LaTeX section drafting, academic text revision, or paper-writing feedback loops to DeepSeek.

Turn a refined research proposal or method idea into a detailed, claim-driven experiment roadmap. Use when the user asks for a detailed experiment plan, ablation matrix, evaluation protocol, run order, compute budget, or paper-ready validation that supports the core problem, novelty, simplicity, and any LLM / VLM / Diffusion / RL-based contribution.

Generate and rank theory-first or proof-oriented research ideas. Use when the user wants non-experimental research ideas, theoretical methods, proof programs, theorem candidates, impossibility results, convergence/sample-complexity analyses, or "analysis" variants of idea creation.

Orchestrate a full theory-first idea discovery pipeline. Use when the user wants the analysis counterpart of idea-discovery: literature context, proof-oriented idea generation, novelty checking, DeepSeek critique, and an analysis plan instead of an experiment-first workflow.

Chrome-dependent academic literature scouting and idea planning. Use when the user gives a research idea, topic, question, or draft direction and wants Codex to browse Google, Google Scholar, and academic sources through the Codex Chrome plugin, identify relevant papers, explain what can be borrowed from them, and write a planNN.md before collecting PDFs or BibTeX.

Chrome-dependent academic paper fetching and BibTeX curation. Use when the user provides a planNN.md, paper list, citation list, or literature plan and wants Codex to use Google, Google Scholar, publisher pages, DBLP, arXiv, and the user's Chrome-accessible permissions to fetch the best available PDFs and create or update references.bib.

Convert configured local Markdown collections to PDFs and upload the generated PDFs to Feishu/Lark Drive. Use when Codex needs to run or maintain a recurring local Markdown-to-PDF publishing workflow, inspect a private rules file, dry-run source/category mappings, invoke an external open-source Markdown PDF converter, or upload generated PDFs to a Feishu Drive folder.

21-rule prose review workflow for technical and academic Markdown/LaTeX excerpts. Use when the user asks for muxing-style-review, wants a complete agent-style check combining check-human.md with agent-style audit-only evidence, wants a compact-rule check, wants a DeepSeek/Codex manuscript draft checked for clarity, evidence, citation discipline, terminology, AI-tell prose, or wants an A/B draft comparison. Prefer a dedicated review subagent when available for non-trivial reviews. Use the sibl...

Verify research idea novelty against recent literature. Use when user says "查新", "novelty check", "有没有人做过", "check novelty", or wants to verify a research idea is novel before implementing.

Independent DeepSeek-backed adversarial proof-audit step for existing theorem, lemma, proposition, or proof artifacts in Markdown, LaTeX, or proof logs. Use when asked to check, audit, verify, red-team, or adversarially review a proof; when a completed proof task needs a correctness pass; or when a broader proof workflow dispatches an independent reviewer to find gaps, hidden assumptions, counterexamples, or unjustified steps.

Synthesize source-indexed proof-material items into the final proof-usage playbook. Use when Codex needs to cluster proof items by proof shape, de-duplicate across sources, design reusable taxonomy entries, update proof-usage indexes and source maps, or preserve material-ID links from cooked proof entries.

Mine proof-heavy papers, notes, PDFs, Markdown, or LaTeX sources into source-indexed proof-material files. Use when the user wants to extract technically nontrivial lemmas, estimates, definitions, dependencies, reductions, constructions, or proof strategies, preserve paper locations and stable material IDs, run DeepSeek screening/backtests, and update the proof-material index rather than writing directly to proof-usage.

Store and inspect source-indexed proof material extracted from papers, notes, PDFs, or LaTeX sources. Use when Codex needs stable material IDs, source locations, reusable abstractions, proof skeletons, tags, and public attribution before any proof ideas are cooked into proof-usage.

Coordinate proof runs across three interfaces: exhaust actionable local proof work first, package exact unresolved obligations for a provider-neutral external solver only when justified, and keep the user informed and in control of consequential choices.

Prepare GPT-pro-ready proof planning or diagnosis bundles from a user's idea, packaging claims, sources, givens, and open questions for review without solving gaps locally.

Look up and apply the cooked, indexed proof playbook built from proof-material. Use when the user is proving a theorem, planning a proof, seeking reusable proof moves by task shape, assumptions, tags, desired move, or material source ID, or adapting cooked local moves and macro strategies with public source attribution.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Codex's capabilities with specialized knowledge, workflows, or tool integrations.

Install or update Codex skills into $CODEX_HOME/skills from a curated list or a GitHub repo path, then surface any per-skill install.md guidance. Use when a user asks to list installable skills, install a curated skill, update an installed skill, or install a skill from another repo (including private repos).

Record observable skill test runs for debugging. Use when the user wants to test, dry-run, debug, inspect, or compare how a Codex skill behaves on a concrete request; creates a local non-overwriting Markdown report with the visible process, files read, commands run, changed artifacts, errors, and a brief behavior summary. Does not certify skill quality or replace user judgment.

Review Markdown prose against agent-style's 21 rules. `/style-review FILE` audits the file and (on user confirm) writes a polished copy at `FILE.reviewed.md`. `/style-review A.md B.md` A/B-compares two drafts and emits a per-rule delta table. Complements the generation-time rules installed via `agent-style enable <tool>`.

Manage, inventory, standardize, migrate, and test reusable Codex skills for the MetaSkill repository. Use when Codex needs to inspect local skills, convert a personal workflow into a public skill candidate, normalize skill folder shape, plan skill promotion, or decide whether a skill belongs in skill-management, publication-review, or skill-examples.