[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7FDZFRRMBW4WZR3FFQ4T8YS":3,"skill-same:01M7FDZFRRMBW4WZR3FFQ4T8YS":48},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":26,"commitSha":27,"skillDirectory":28,"forksCount":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M7FDZFRRMBW4WZR3FFQ4T8YS","proof-checker",null,"Rigorous mathematical proof verification and fixing workflow. Reads a LaTeX proof, identifies gaps via fresh-agent Codex GPT-6-Astra ultra review, fixes each gap with full derivations, re-reviews, and generates an audit report. Base review is same-family provisional. Use when user says \"检查证明\", \"verify proof\", \"proof check\", \"审证明\", \"check this proof\", or wants rigorous mathematical verification of a theory paper.","CZX-LBJ\u002Fdeepseek-science","CZX-LBJ","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F204594689?v=4","01M7FC59WQ9PFXT04MKGJDCFFT","czx","01M7FC59XB2GW0BKWN9JTDBQS7",1,"MIT","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fskills-codex\u002Fproof-checker\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: proof-checker\ndescription: Rigorous mathematical proof verification and fixing workflow. Reads a LaTeX proof, identifies gaps via fresh-agent Codex GPT-6-Astra ultra review, fixes each gap with full derivations, re-reviews, and generates an audit report. Base review is same-family provisional. Use when user says \"检查证明\", \"verify proof\", \"proof check\", \"审证明\", \"check this proof\", or wants rigorous mathematical verification of a theory paper.\nargument-hint: \"[path-to-tex-file or proof-description]\"\nallowed-tools: Bash(*), Read, Grep, Glob, Write, Edit\n---\n\n# Proof Checker: Rigorous Mathematical Verification & Fixing\n\n> **Codex assurance:** base Codex proof judgments are\n> `review_independence: same-family` and `acceptance_status: provisional`.\n> Deterministic compilation\u002Falgebra checks may be accepted; a semantic proof\n> acceptance requires a cross-family overlay. Reviewer failure emits BLOCKED.\n\nSystematically verify a mathematical proof via fresh-agent adversarial review, fix identified gaps, re-review until convergence, and generate a detailed audit report with proof-obligation accounting.\n\n## Context: $ARGUMENTS\n\n## Constants\n\n- MAX_REVIEW_ROUNDS = 3\n- REVIEWER_MODEL = `gpt-6-astra` via Codex reviewer agent, reasoning effort\n  `ultra` for this deep-audit skill (capability fallback never below `xhigh`)\n- **REVIEWER_BACKEND = `codex`** — Default: Codex reviewer agent (`spawn_agent`, ultra — deep-audit tier). Override with `— reviewer: oracle-pro` for GPT-5.5 Pro via Oracle MCP. See `shared-references\u002Freviewer-routing.md`.\n- AUDIT_DOC: `PROOF_AUDIT.md` at the paper directory root, alongside `main.tex` (cumulative log; when invoked via `\u002Fpaper-writing`, this is `paper\u002FPROOF_AUDIT.md`)\n- REPORT_TEX: `proof_audit_report.tex` (formal before\u002Fafter PDF)\n- STATE_FILE: `PROOF_CHECK_STATE.json` (for recovery)\n- SKELETON_DOC: `PROOF_SKELETON.md` (micro-claim inventory)\n- **RENDER_HTML = true** — When `true` (default), auto-render `PROOF_AUDIT.md` to HTML at workflow end via `\u002Frender-html`. Uses **full review gate** (audit-class, math-heavy — render-fidelity check protects against MathJax breakage). Set `false` to skip, or pass `— render html: false`. **Non-blocking**: failures don't invalidate the proof audit.\n\n### Acceptance Gate (objective, replaces subjective scoring)\n\nThe proof passes when ALL of the following hold:\n1. Zero open FATAL or CRITICAL issues\n2. Every theorem\u002Flemma has: (i) explicit hypotheses, (ii) proof with all interchanges justified, (iii) every application discharges hypotheses in the ledger\n3. All big-O\u002FΘ\u002Fo statements have declared parameter dependence and uniformity scope\n4. Counterexample pass executed on all key lemmas (log candidates even if none found)\n\n## Issue Taxonomy (20 categories, 4 groups)\n\n### Group A: Logic & Proof Structure\n\n| Category | Description | Example |\n|----------|-------------|---------|\n| **UNJUSTIFIED_ASSERTION** | Claim stated without proof or reference | \"The Hessian splits into Gram blocks\" |\n| **UNPROVEN_SUBCLAIM** | \"Clearly\" \u002F \"it follows\" hides a nontrivial lemma | \"By symmetry, the cross-terms vanish\" without checking |\n| **QUANTIFIER_ERROR** | Wrong order ∀\u002F∃, missing \"for sufficiently small κ\" | \"For all π, there exists ε\" vs \"there exists ε for all π\" |\n| **IMPLICATION_REVERSAL** | Uses (A⇒B) as (B⇒A), or claims equivalence with only one direction | |\n| **CASE_INCOMPLETE** | Misses boundary\u002Fdegenerate cases | Singular covariance, zero weight, non-unique argmin |\n| **CIRCULAR_DEPENDENCY** | Lemma uses theorem that depends on it | |\n| **LOGICAL_GAP** | A step is not justified by what precedes it | B=Θ(1) → β_K=0 without analyzing W |\n\n### Group B: Analysis & Measure Theory\n\n| Category | Description | Example |\n|----------|-------------|---------|\n| **ILLEGAL_INTERCHANGE** | Swaps limit\u002Fexpectation\u002Fderivative\u002Fintegral without DCT\u002FMCT\u002FFubini | Differentiating under E without domination |\n| **NONUNIFORM_CONVERGENCE** | Pointwise convergence used as uniform | sup and limit swapped |\n| **MISSING_DOMINATION** | DCT cited but no dominating function given | |\n| **INTEGRABILITY_GAP** | Uses E|X|^p without proving\u002Fassuming finite moments | |\n| **REGULARITY_GAP** | Differentiability\u002FLipschitz\u002Fconvexity used but not established | |\n| **STOCHASTIC_MODE_CONFUSION** | Mixes a.s.\u002Fin prob.\u002Fin L²\u002Fin expectation | |\n\n### Group C: Model & Parameter Tracking\n\n| Category | Description | Example |\n|----------|-------------|---------|\n| **MISSING_DERIVATION** | A quantity is used but never derived from the model | Risk functional with undefined B, W |\n| **HIDDEN_ASSUMPTION** | Proof silently uses a condition not in the theorem | Gaussianity assumed but not stated |\n| **INSUFFICIENT_ASSUMPTION** | Hypotheses too weak for proof (counterexample exists) | Moment conditions admitting 2-point distributions |\n| **DIMENSION_TRACKING** | Parameter dependence (d, n, K, ...) not explicit | d enters only through κ |\n| **NORMALIZATION_MISMATCH** | Coordinate\u002Fscaling conventions inconsistent | Rescaled vs raw coordinates |\n| **CONSTANT_DEPENDENCE_HIDDEN** | \"C\" depends on d,n,K but treated as universal | |\n\n### Group D: Scope & Claims\n\n| Category | Description | Example |\n|----------|-------------|---------|\n| **SCOPE_OVERCLAIM** | Conclusion stated more broadly than proof supports | \"β_K=0\" with only generic overlap |\n| **REFERENCE_MISMATCH** | Cited theorem's hypotheses not verified at point of use | |\n\n## Two-Axis Severity System\n\n### Axis A — Proof Status (what is wrong)\n\n| Status | Meaning |\n|--------|---------|\n| **INVALID** | Statement false as written (counterexample exists or contradiction) |\n| **UNJUSTIFIED** | Could be true, but current proof does not establish it |\n| **UNDERSTATED** | True only after strengthening assumptions |\n| **OVERSTATED** | True only after weakening conclusion \u002F adding qualifiers |\n| **UNCLEAR** | Ambiguous notation \u002F definition drift (not wrong per se) |\n\n### Axis B — Impact (how much breaks)\n\n| Impact | Meaning |\n|--------|---------|\n| **GLOBAL** | Breaks main theorem or core dependency chain |\n| **LOCAL** | Affects a side result but not the main theorem |\n| **COSMETIC** | Exposition only |\n\n### Severity Labels (derived)\n\n| Label | Definition |\n|-------|------------|\n| **FATAL** | INVALID + GLOBAL |\n| **CRITICAL** | (INVALID + LOCAL) or (UNJUSTIFIED + GLOBAL) |\n| **MAJOR** | (UNJUSTIFIED + LOCAL) or (UNDERSTATED\u002FOVERSTATED + GLOBAL) |\n| **MINOR** | Clarity \u002F notation \u002F dimension bookkeeping that doesn't change claims |\n\n## Side-Condition Checklists for Common Theorems\n\nWhen the proof invokes any of the following, require explicit verification of ALL listed conditions:\n\n| Theorem | Required Conditions |\n|---------|-------------------|\n| **DCT** (Dominated Convergence) | Pointwise a.e. convergence + integrable dominating function |\n| **MCT** (Monotone Convergence) | Monotone increasing + non-negative |\n| **Fubini\u002FTonelli** | Product measurability + integrability (Fubini) or non-negative (Tonelli) |\n| **Leibniz integral rule** | Continuity of integrand + dominating function for derivative |\n| **Implicit Function Theorem** | Continuous differentiability + non-singular Jacobian |\n| **Taylor with remainder** | Sufficient differentiability + remainder form (Lagrange\u002Fintegral) |\n| **Jensen's inequality** | Convexity of function + integrability |\n| **Cauchy-Schwarz** | Correct inner product space + integrability of both factors |\n| **Weyl\u002FDavis-Kahan** | Symmetry\u002FHermiticity + perturbation bound conditions |\n| **Analytic continuation** | Domain connectivity + identity theorem conditions |\n| **WLOG reduction** | Invariance under claimed symmetry + reduction is reversible |\n\n## Workflow\n\n### Proof-obligation fan-out\n\nIndependent sections\u002Ftheorems may be extracted by fresh read-only\n`spawn_agent` shards, with a sequential fresh-context fallback. Each shard\nreturns `{\"shard_id\": ..., \"entries\": [{\"payload\": ..., \"dedup_key\":\n\"\u003Ctheorem-or-obligation-id>\"}]}` and must not declare the proof valid. The\nparent mechanically merges obligations; the fresh Codex review that evaluates\nthem records `review_independence: same-family` and\n`acceptance_status: provisional`. See\n[`fan-out-pattern.md`](..\u002Fshared-references\u002Ffan-out-pattern.md).\n\n### Phase 0: Preparation\n\n1. **Locate the proof**: Find the main `.tex` file(s).\n2. **Read the entire proof**: Extract list of all theorems\u002Flemmas\u002Fpropositions\u002Fcorollaries\u002Fdefinitions\u002Fassumptions.\n3. **Read reference materials**: Reference papers, prior results.\n4. **Build a section map**: Structured list with line numbers and key claims.\n5. **Identify the main theorem**: Central result, assumptions, claims.\n\n### Phase 0.5: Proof-Obligation Ledger\n\nBuild formal accounting artifacts. Save to `PROOF_SKELETON.md`:\n\n#### 1. Dependency DAG\nNodes = Definitions \u002F Assumptions \u002F Lemmas \u002F Theorems. Edges = \"uses\". **Detect cycles** (including semantic circularity where Lemma A uses a corollary that quietly depends on A).\n\n#### 2. Assumption Ledger\nFor each theorem\u002Flemma, list every hypothesis with WHERE each is verified (or mark \"UNVERIFIED\"). Track **usage-minimal assumption sets** — which assumptions were actually used vs merely stated.\n\n#### 3. Typed Symbol Table\nEach symbol must have a **type signature**:\n```\nκ : scalar ∈ (0,1), depends on (d, α_t, Σ, μ)\nu* : vector ∈ ℝ^d, u* = C^{-1}m\nB^even : matrix ∈ ℝ^{(L+1)×(L+1)}, symmetric PSD\nΨ_v : function ℝ → ℝ, analytic in (ζ,κ), parity determined by v\n```\nFlag any symbol whose meaning changes or whose type is inconsistent across uses.\n\n#### 4. Canonical Quantified Statements\nFor each theorem\u002Flemma, rewrite the statement with **explicit quantifiers, domains, and limit order**:\n```\n∀K ≥ 3, ∀π ∈ Π_K^{ms,∘} \\ E_K, ∃κ_0 > 0 such that ∀κ ∈ (0, κ_0):\n  h_act^{(K,π)} = Θ(κ^{α_K^act})  [uniform in π on compact subsets]\n```\nIf you cannot restate a theorem this precisely, mark it **UNCLEAR — needs disambiguation**.\n\n#### 5. Micro-Claim Inventory\nEvery nontrivial step becomes a numbered micro-claim in **sequent form**:\n```\nMC-17: Context: [Lemma 3.1, κ \u003C κ_0, Z_κ has bounded moments up to order 2m+2]\n       ⊢ Goal: P̂_0 is positive definite\n       Rule: monomials linearly independent on support of continuous distribution\n       Side-conditions: positive density near origin ✓ (by GMM weak convergence)\n```\nEach micro-claim has: justification rule name + required conditions + where conditions are proven.\n\n#### 6. Limit-Order Map\nTrack every asymptotic statement's **limit order and uniformity scope**:\n```\nh_act = Θ(κ^α)  [as κ→0, uniform in π on compact subsets of Π_K, for fixed K]\nτ_act ~ (b\u002Fa)n   [as n→∞, for fixed κ,K,π with x_K ≪ 1]\n```\nFlag any statement where limit order is ambiguous or uniformity is unclear.\n\n### Optional Lean verification\n\nWhen the user requests Lean or a specific obligation warrants formal checking,\ndelegate that obligation and its original hypotheses to\n[`\u002Flean-formalize`](..\u002Flean-formalize\u002FSKILL.md). Record the checked scope, exported\ndeclaration, statement-alignment evidence, build output and transitive-axiom\naudit, then resume this audit. A proved sublemma closes only the corresponding\nobligation. Keep the complete proof audit, acceptance decision and canonical wiki\nartifacts here; the nested Lean task returns evidence without starting another\n`proof-checker` run. Routine proof checks do not require Lean. Reviewer agreement\nand unsuccessful counterexample searches do not discharge a proof obligation.\n\n### Phase 1: First Review (Codex GPT-6-Astra ultra)\n\nSubmit the **complete proof content** with the following **mandatory reviewer checklist** in the prompt:\n\n```text\nspawn_agent:\n  model: gpt-6-astra\n  reasoning_effort: ultra\n  message: |\n    You are performing a rigorous mathematical proof review. For EVERY theorem,\n    lemma, and proposition, check ALL of the following:\n\n    ## MANDATORY CHECKS\n\n    A. DEFINITIONS: List any symbol whose meaning is ambiguous or changes.\n    B. HYPOTHESIS DISCHARGE: For each lemma\u002Ftheorem APPLICATION (not statement),\n       list each hypothesis and whether it was verified, with location.\n    C. INEQUALITY AUDIT: For each inequality chain, verify direction, missing\n       absolute values, missing conditions (convexity, PSD, integrability).\n    D. INTERCHANGE AUDIT: Flag every limit\u002Fderivative\u002Fexpectation\u002Fintegral\n       interchange. State which theorem justifies it (DCT\u002FMCT\u002FFubini\u002FLeibniz)\n       and which conditions are verified\u002Fmissing.\n    E. PROBABILITY MODE: Track whether claims are a.s.\u002Fin prob.\u002Fin expectation\u002F\n       w.h.p. Ensure transitions are justified.\n    F. UNIFORMITY & CONSTANTS: For every O(·), o(·), Θ(·), ≲, state whether\n       it is uniform over all parameters. List hidden parameter dependence.\n    G. EDGE\u002FDEGENERATE CASES: Attempt to break each key lemma with a 1D,\n       low-rank, or extreme-parameter construction.\n    H. DEPENDENCY CONSISTENCY: Detect cycles or forward references to unproven\n       results.\n\n    ## OUTPUT FORMAT (per issue)\n    For each issue found, provide:\n    - id: sequential number\n    - status: INVALID \u002F UNJUSTIFIED \u002F UNDERSTATED \u002F OVERSTATED \u002F UNCLEAR\n    - impact: GLOBAL \u002F LOCAL \u002F COSMETIC\n    - category: [from taxonomy]\n    - location: section\u002Fequation\u002Fline\n    - statement: what the proof claims\n    - why_invalid: why this is wrong or unjustified\n    - counterexample: YES (describe) \u002F NO \u002F CANDIDATE (describe attempt)\n    - affects: which downstream results break if this is wrong\n    - minimal_fix: how to fix it\n\n    [FULL PROOF CONTENT HERE]\n```\n\n**Save the reviewer `agent_id`.** Parse into structured issue list. Write to `PROOF_AUDIT.md`.\n\n### Phase 1.5: Counterexample Red Team\n\nFor each CRITICAL or MAJOR issue, and for every key lemma that introduces:\n- a new inequality bound\n- an identifiability\u002Funiqueness claim\n- a curvature\u002FPSD\u002Fstrong convexity assertion\n- a uniform-in-parameter claim\n- a convergence mode upgrade (pointwise → uniform, in prob → w.h.p.)\n\nSystematically attempt to construct counterexamples using:\n\n| Strategy | Description |\n|----------|-------------|\n| **Dimensional collapse** | Set d=1 or 2, K=2, n small |\n| **Degeneracy** | Singular covariance, tiny weight, overlapping means, identical components |\n| **Extremal distributions** | Two-point ±a, bounded non-subGaussian, heavy tails |\n| **Adversarial parameter scaling** | Pick parameters making neglected terms dominate |\n| **Numeric falsification** | Translate lemma to a function, brute-force optimize over small domain |\n\n**Rule**: Label \"counterexample found\" ONLY if algebraically verified. Otherwise log as \"candidate counterexample — needs verification.\"\n\nRecord all attempts (successful or not) in `PROOF_AUDIT.md`.\n\n### Phase 2: Fix Implementation\n\nFor each issue, ordered by severity (FATAL → CRITICAL → MAJOR → MINOR):\n\n#### Step 2a: Choose fix strategy\nFor each issue, explicitly choose one of:\n- **ADD_DERIVATION**: Write missing proof steps\n- **STRENGTHEN_ASSUMPTION**: Add conditions to theorem statement\n- **WEAKEN_CLAIM**: Reduce conclusion scope\n- **ADD_REFERENCE**: Cite known result + verify its conditions apply\n\nLog this choice — it is a scope-changing decision when it alters theorem statements.\n\n#### Step 2b: Derive the fix mathematically\n- Complete mathematical derivation, not just a claim\n- If new proposition\u002Flemma needed, write in full theorem-proof style\n\n#### Step 2c: Implement in LaTeX\n- Edit the `.tex` file\n- Preserve existing `\\label` references where possible\n\n#### Step 2d: Record the fix\n```markdown\n### Fix N: [SHORT TITLE]\n**Issue**: [id] [CATEGORY] — [description]\n**Severity**: FATAL \u002F CRITICAL \u002F MAJOR \u002F MINOR\n**Status**: INVALID \u002F UNJUSTIFIED \u002F UNDERSTATED \u002F OVERSTATED\n**Impact**: GLOBAL \u002F LOCAL \u002F COSMETIC\n**Fix strategy**: ADD_DERIVATION \u002F STRENGTHEN_ASSUMPTION \u002F WEAKEN_CLAIM \u002F ADD_REFERENCE\n**Location**: Section X, Lines Y-Z\n\n**BEFORE**: [what the proof originally did]\n**WHY WRONG**: [mathematical problem, with counterexample if applicable]\n**AFTER**: [what the fix does]\n**KEY EQUATION**: [central new equation]\n**PROOF OBLIGATIONS ADDED**: [new conditions\u002Flemmas introduced]\n**DOWNSTREAM EFFECTS**: [which results now need re-checking]\n```\n\n#### Step 2e: Compile check\n```bash\npdflatex -interaction=nonstopmode \u003Cfile>.tex 2>&1 | grep -E \"Error|Warning|undefined\"\n```\n\n### Phase 3: Re-Review (Codex GPT-6-Astra ultra)\n\nLaunch a fresh reviewer agent for the next review round. Do not use `send_input` here; proof-checker keeps each round independent. Request the same mandatory checklist.\n\nCheck acceptance gate. If not met, repeat Phases 2-3 (up to MAX_REVIEW_ROUNDS).\n\n### Phase 3.5: Global Closure & Independent Verification\n\n#### Global closure checks\nAfter all fixes, verify the proof as a whole:\n- **Statement–conclusion match**: Does the proof end with EXACTLY what the theorem claims (quantifiers, constants, uniformity)?\n- **All obligations discharged**: Every node in the obligation DAG is proven or explicitly assumed (and the theorem statement includes it).\n- **Case analysis coverage**: Cases partition the domain AND include boundary\u002Fdegenerate cases.\n- **Induction correctness** (if applicable): Base case, inductive step, correct use of IH, induction measure strictly decreases.\n- **WLOG reductions**: Each \"without loss of generality\" spawns a micro-claim proving the reduction is lossless.\n- **No silent assumption strengthening**: Any fix that strengthened assumptions has propagated to the main theorem statement.\n\n#### Independent second review for FATAL\u002FCRITICAL fixes\nFor any fix that resolved a FATAL or CRITICAL issue, submit the **fixed section alone** (without showing the previous critique) to a **fresh Codex thread**:\n\n```text\nspawn_agent:\n  model: gpt-6-astra\n  reasoning_effort: ultra\n  message: |\n    Blind review of the following proof section. You have NOT seen any prior\n    review or discussion. Check every step for correctness, hidden assumptions,\n    illegal interchanges, and counterexamples.\n    [FIXED SECTION ONLY]\n```\n\nIf the blind reviewer finds new issues, re-enter Phase 2.\n\n#### Regression proof-audit\nAfter fixes, re-run:\n- DAG acyclicity check (no new cycles introduced)\n- Counterexample suite on all DOWNSTREAM lemmas of modified results\n- Assumption-delta report: what became stronger\u002Fweaker due to fixes?\n\n### Phase 3.9: Unrecoverable Proof Protocol\n\nIf acceptance gate is not met after MAX_REVIEW_ROUNDS, output a **Proof Unrecoverable Report**:\n1. Minimal set of blocking FATAL\u002FCRITICAL issues that could not be resolved\n2. Salvage options ranked: (a) weaken claim, (b) strengthen assumptions, (c) add missing lemmas, (d) restructure argument\n3. Which parts of the proof are likely still reusable\n4. Recommended next steps for the author\n\nDo NOT silently declare success. The report must be honest.\n\n### Phase 4: Audit Report Generation\n\nGenerate `proof_audit_report.tex` with:\n\n1. **Overview table**: All issues with two-axis severity, category, fix strategy, status\n2. **Before\u002FAfter logic chain**: Red (BEFORE) → Green (AFTER) comparison\n3. **For each fix**: original proof → why wrong → counterexample (if any) → complete derivation → remaining subtleties\n4. **Proof-obligation diff**: What was unverified before, what is verified now\n5. **Summary**: Now proven \u002F still assumed \u002F open problems\n6. **Colored boxes**: BEFORE (red), AFTER (green), WHY WRONG (orange), KEY INSIGHT (blue), WARNING (yellow)\n\nCompile: `pdflatex proof_audit_report.tex && pdflatex proof_audit_report.tex`\n\n### Phase 5: State Persistence\n\nWrite `PROOF_CHECK_STATE.json`:\n```json\n{\n  \"status\": \"completed\",\n  \"rounds\": 2,\n  \"review_agent_ids\": [\"...\"],\n  \"fatal_fixed\": 0,\n  \"critical_fixed\": 3,\n  \"major_fixed\": 2,\n  \"minor_fixed\": 1,\n  \"counterexamples_found\": 1,\n  \"counterexample_candidates\": 2,\n  \"acceptance_gate\": \"PASS\",\n  \"timestamp\": \"...\"\n}\n```\n\n### Phase 5.5: Research Wiki Claim Ledger (additive; only if a wiki is active)\n\nIf — and only if — a `research-wiki\u002F` exists, persist each top-level theorem\u002Fheadline\nas a **claim node** (the wiki's PROVE\u002FJUDGE ledger). This is the **birth point** for wiki\nclaim nodes. It is a **detect-only record, never a verdict**: it never changes the audit's\n`verdict`\u002F`reason_code`, never blocks, and is skipped when `verdict == NOT_APPLICABLE` or no\nwiki is found. The claim's `status` is the **PROOF axis only** ({drafted, unproven,\nsound-modulo-imports, verified, refuted, retracted}); empirical experiment support is a\nseparate axis carried by edges (`\u002Fresult-to-claim`), never written into `status`.\n\nResolve the helper via the Codex-side chain (skip cleanly if unavailable; the audit is\nalready complete):\n```\nARIS_REPO=\"${ARIS_REPO:-$(awk -F'\\t' '$1==\"repo_root\"{print $2; exit}' .aris\u002Finstalled-skills-codex.txt 2>\u002Fdev\u002Fnull)}\"\nWIKI_SCRIPT=\"\"\n[ -n \"$ARIS_REPO\" ] && [ -f \"$ARIS_REPO\u002Ftools\u002Fresearch_wiki.py\" ] && WIKI_SCRIPT=\"$ARIS_REPO\u002Ftools\u002Fresearch_wiki.py\"\n[ -z \"$WIKI_SCRIPT\" ] && [ -f tools\u002Fresearch_wiki.py ] && WIKI_SCRIPT=\"tools\u002Fresearch_wiki.py\"\n[ -z \"$WIKI_SCRIPT\" ] && [ -f ~\u002F.codex\u002Fskills\u002Fresearch-wiki\u002Fresearch_wiki.py ] && WIKI_SCRIPT=\"$HOME\u002F.codex\u002Fskills\u002Fresearch-wiki\u002Fresearch_wiki.py\"\n```\n\nIf `research-wiki\u002F` exists and `WIKI_SCRIPT` is available and `verdict != NOT_APPLICABLE`,\nfor each top-level theorem map the audit outcome to an honest status — `PASS`\u002Fall proofs\ncomplete → `verified`; closes modulo flagged imports → `sound-modulo-imports`; counterexample\nfound or statement judged false → `refuted`; open gap (UNJUSTIFIED, no counterexample) →\n`unproven` (never fake a gap as `refuted`\u002F`verified`) — then record it (idempotent):\n```\npython3 \"$WIKI_SCRIPT\" add_claim research-wiki\u002F --slug \"\u003Cstable-theorem-id>\" \\\n     --name \"\u003Ctheorem headline>\" --status \"\u003Cmapped status>\" \\\n     --provenance \"\u003Ctrace_path from PROOF_AUDIT.json>\" --statement \"\u003Ccanonical statement>\" \\\n     --scope \"\u003Cwhat it does NOT say; flagged imports>\" --update-on-exist\n```\n`add_claim` failure is non-fatal (warn and continue; the audit is unaffected).\n\n## Key Rules\n\n### Mathematical rigor\n- **Never accept a proof step on faith**. \"Clearly\" \u002F \"it follows\" \u002F \"by standard arguments\" are red flags — each must spawn a micro-claim.\n- **Hypothesis discharge**: Every time a lemma is APPLIED, verify EACH of its hypotheses at that point. Use the side-condition checklists above.\n- **Interchange discipline**: Every swap of limit\u002Fexpectation\u002Fderivative\u002Fintegral must cite a theorem (DCT\u002FMCT\u002FFubini\u002FLeibniz) and verify its conditions with explicit dominating function or integrability proof.\n- **Uniformity discipline**: Every O(·)\u002FΘ(·) must declare what parameters it is uniform over. \"O(1)\" that secretly depends on d,n,K is a CONSTANT_DEPENDENCE_HIDDEN issue.\n- **Quantifier discipline**: Check ∀\u002F∃ order. \"For sufficiently small κ\" must specify: does κ₀ depend on K? On π? On d?\n- **Counterexample-first**: Before trying to fix a gap, first try to break it.\n- **WLOG prohibition**: Every \"without loss of generality\" must have an explicit micro-claim proving the reduction. No free WLOGs.\n- **No silent assumption strengthening**: Any fix that adds conditions must propagate to the theorem statement.\n\n### Review-independence protocol\n- **Codex executor analyzes and implements; a fresh Codex reviewer provides adversarial review.** Base review remains same-family\u002Fprovisional.\n- **Codex reasoning always ultra** (deep-audit tier): never below `xhigh` — only the capability fallback in `reviewer-routing.md` may step down, and only on explicit capability errors.\n- **Send full content**: Don't summarize — send actual math for line-by-line checking.\n- **Fresh reviewer agents**: Save each returned `agent_id` for traceability, but launch a new `spawn_agent` for each review round. Do not use `send_input` across proof-checker rounds.\n\n### Fix quality\n- **Minimal fixes**: Fix exactly what's broken, nothing more.\n- **Full derivation**: Every fix includes complete mathematical argument.\n- **Explicit scope decisions**: Each fix is tagged ADD_DERIVATION \u002F STRENGTHEN_ASSUMPTION \u002F WEAKEN_CLAIM \u002F ADD_REFERENCE.\n- **Compile after each fix**: LaTeX must compile cleanly.\n\n### Scope honesty\n- **Don't overclaim**: If a fix makes a result conditional, say so.\n- **Separate \"proven\" from \"assumed\"**: The audit report has an explicit section for this.\n- **Log open problems**: Issues requiring future work are listed, not hidden.\n\n## Output Files\n\n| File | Content | When |\n|------|---------|------|\n| `PROOF_SKELETON.md` | Dependency DAG + assumption ledger + micro-claims | Phase 0.5 |\n| `PROOF_AUDIT.md` | Cumulative round-by-round audit log | Updated each round |\n| `PROOF_AUDIT.json` | Machine-readable submission verdict (see below) | Always emitted |\n| `proof_audit_report.tex\u002F.pdf` | Formal before\u002Fafter report | Phase 4 |\n| `PROOF_CHECK_STATE.json` | State for recovery | Phase 5 |\n| `PROOF_AUDIT.html` (+ `.review.json` sidecar) | Single-file HTML view auto-rendered via `\u002Frender-html \"PROOF_AUDIT.md\" --json \"PROOF_AUDIT.json\"`. **Non-blocking** — if `\u002Frender-html` fails the audit still counts as complete. | Workflow end (when `RENDER_HTML = true`, default) |\n\n## Submission Artifact Emission\n\nThis skill **always** writes `PROOF_AUDIT.json` at the paper directory\nroot (i.e. `paper\u002FPROOF_AUDIT.json` when invoked from `\u002Fpaper-writing`\nwith paper-dir `paper\u002F`; `\u003Cyour-paper-dir>\u002FPROOF_AUDIT.json` when invoked\nstandalone), regardless of caller or whether the paper contains theorems.\nA paper with no `\\begin{theorem}` \u002F `\\begin{lemma}` \u002F `\\begin{proof}` emits\nverdict `NOT_APPLICABLE`; silent skip is forbidden. `paper-writing`\nPhase 6 and `verify_paper_audits.sh` both rely on this artifact\nexisting at `\u003Cpaper-dir>\u002FPROOF_AUDIT.json`.\n\nThe artifact conforms to the schema in `shared-references\u002Fassurance-contract.md`:\n\n```json\n{\n  \"audit_skill\":      \"proof-checker\",\n  \"verdict\":          \"PASS | WARN | FAIL | NOT_APPLICABLE | BLOCKED | ERROR\",\n  \"reason_code\":      \"all_proofs_complete | minor_gaps | critical_gap | no_theorems | ...\",\n  \"summary\":          \"One-line human-readable verdict summary.\",\n  \"audited_input_hashes\": {\n    \"main.tex\":                 \"sha256:...\",\n    \"sections\u002F4.theory.tex\":    \"sha256:...\"\n  },\n  \"trace_path\":       \".aris\u002Ftraces\u002Fproof-checker\u002F\u003Cdate>_run\u003CNN>\u002F\",\n  \"thread_id\":        \"\u003Ccodex mcp thread id>\",\n  \"executor_model\":   \"codex-gpt-6-astra\",\n  \"executor_family\":  \"openai\",\n  \"reviewer_model\":   \"gpt-6-astra\",\n  \"reviewer_family\":  \"openai\",\n  \"review_independence\": \"same-family\",\n  \"acceptance_status\": \"provisional\",\n  \"reviewer_reasoning\": \"ultra\",\n  \"generated_at\":     \"\u003CUTC ISO-8601>\",\n  \"details\": {\n    \"theorems_audited\": \u003Cint>,\n    \"issues\": [ { \"id\": \"T1-H3\", \"severity\": \"FATAL|CRITICAL|MAJOR|MINOR\",\n                  \"category\": \"quantifier|domination|...\",\n                  \"location\": \"sections\u002F4.theory.tex:L182\",\n                  \"note\": \"...\" }, ... ]\n  }\n}\n```\n\n### `audited_input_hashes` scope\n\nHash the **declared input set** actually reviewed — the theorem-bearing\n`.tex` files passed into this invocation — not a repo-wide union and not\nthe reviewer's self-reported opened subset. The external verifier rehashes\nthese entries; any mismatch flags `STALE`.\n\n**Path convention** (must match `verify_paper_audits.sh`): keys are\n**paths relative to the paper directory** (no `paper\u002F` prefix — the\nverifier resolves relative to the paper dir; prefixing produces\n`paper\u002Fpaper\u002F...` and false-fails as STALE). Use **absolute paths** for\nfiles outside the paper dir.\n\n### Verdict decision table\n\n| Input state                                           | Verdict          | `reason_code` example |\n|-------------------------------------------------------|------------------|-----------------------|\n| No theorems \u002F lemmas \u002F proofs in paper                | `NOT_APPLICABLE` | `no_theorems`         |\n| Theorems present but referenced files unreadable      | `BLOCKED`        | `source_unreadable`   |\n| All proof obligations discharged, no gaps             | `PASS`           | `all_proofs_complete` |\n| Only MINOR issues (notation \u002F exposition)             | `WARN`           | `minor_gaps`          |\n| Any FATAL or CRITICAL issue (logic gap, wrong claim)  | `FAIL`           | `critical_gap`        |\n| Reviewer invocation failed (network \u002F malformed)      | `ERROR`          | `reviewer_error`      |\n\nMAJOR issues alone map to `WARN` or `FAIL` at the reviewer's discretion and\nmust carry an explicit justification in `summary` + `details.issues`.\n\n### Thread independence\n\nEvery invocation uses a fresh reviewer agent. Never use `send_input` across\nproof-checker runs. Do not accept prior audit outputs\n(PAPER_CLAIM_AUDIT, CITATION_AUDIT, EXPERIMENT_LOG) as input — the fresh\nthread preserves reviewer independence per\n`shared-references\u002Freviewer-independence.md`.\n\nThis skill never blocks by itself; `paper-writing` Phase 6 plus the\nverifier decide whether the verdict blocks finalization based on the\n`assurance` level.\n\n## Example Invocations\n\n```\n\u002Fproof-checker \"neurips_2025.tex\"\n\u002Fproof-checker \"check the GMM generalization proof, focus on dimension dependence\"\n\u002Fproof-checker \"verify proof in paper.tex — difficulty: nightmare\"\n```\n","# Proof Checker: Rigorous Mathematical Verification & Fixing\n\n> **Codex assurance:** base Codex proof judgments are\n> `review_independence: same-family` and `acceptance_status: provisional`.\n> Deterministic compilation\u002Falgebra checks may be accepted; a semantic proof\n> acceptance requires a cross-family overlay. Reviewer failure emits BLOCKED.\n\nSystematically verify a mathematical proof via fresh-agent adversarial review, fix identified gaps, re-review until convergence, and generate a detailed audit report with proof-obligation accounting.\n\n## Context: $ARGUMENTS\n\n## Constants\n\n- MAX_REVIEW_ROUNDS = 3\n- REVIEWER_MODEL = `gpt-6-astra` via Codex reviewer agent, reasoning effort\n  `ultra` for this deep-audit skill (capability fallback never below `xhigh`)\n- **REVIEWER_BACKEND = `codex`** — Default: Codex reviewer agent (`spawn_agent`, ultra — deep-audit tier). Override with `— reviewer: oracle-pro` for GPT-5.5 Pro via Oracle MCP. See `shared-references\u002Freviewer-routing.md`.\n- AUDIT_DOC: `PROOF_AUDIT.md` at the paper directory root, alongside `main.tex` (cumulative log; when invoked via `\u002Fpaper-writing`, this is `paper\u002FPROOF_AUDIT.md`)\n- REPORT_TEX: `proof_audit_report.tex` (formal before\u002Fafter PDF)\n- STATE_FILE: `PROOF_CHECK_STATE.json` (for recovery)\n- SKELETON_DOC: `PROOF_SKELETON.md` (micro-claim inventory)\n- **RENDER_HTML = true** — When `true` (default), auto-render `PROOF_AUDIT.md` to HTML at workflow end via `\u002Frender-html`. Uses **full review gate** (audit-class, math-heavy — render-fidelity check protects against MathJax breakage). Set `false` to skip, or pass `— render html: false`. **Non-blocking**: failures don't invalidate the proof audit.\n\n### Acceptance Gate (objective, replaces subjective scoring)\n\nThe proof passes when ALL of the following hold:\n1. Zero open FATAL or CRITICAL issues\n2. Every theorem\u002Flemma has: (i) explicit hypotheses, (ii) proof with all interchanges justified, (iii) every application discharges hypotheses in the ledger\n3. All big-O\u002FΘ\u002Fo statements have declared parameter dependence and uniformity scope\n4. Counterexample pass executed on all key lemmas (log candidates even if none found)\n\n## Issue Taxonomy (20 categories, 4 groups)\n\n### Group A: Logic & Proof Structure\n\n| Category | Description | Example |\n|----------|-------------|---------|\n| **UNJUSTIFIED_ASSERTION** | Claim stated without proof or reference | \"The Hessian splits into Gram blocks\" |\n| **UNPROVEN_SUBCLAIM** | \"Clearly\" \u002F \"it follows\" hides a nontrivial lemma | \"By symmetry, the cross-terms vanish\" without checking |\n| **QUANTIFIER_ERROR** | Wrong order ∀\u002F∃, missing \"for sufficiently small κ\" | \"For all π, there exists ε\" vs \"there exists ε for all π\" |\n| **IMPLICATION_REVERSAL** | Uses (A⇒B) as (B⇒A), or claims equivalence with only one direction | |\n| **CASE_INCOMPLETE** | Misses boundary\u002Fdegenerate cases | Singular covariance, zero weight, non-unique argmin |\n| **CIRCULAR_DEPENDENCY** | Lemma uses theorem that depends on it | |\n| **LOGICAL_GAP** | A step is not justified by what precedes it | B=Θ(1) → β_K=0 without analyzing W |\n\n### Group B: Analysis & Measure Theory\n\n| Category | Description | Example |\n|----------|-------------|---------|\n| **ILLEGAL_INTERCHANGE** | Swaps limit\u002Fexpectation\u002Fderivative\u002Fintegral without DCT\u002FMCT\u002FFubini | Differentiating under E without domination |\n| **NONUNIFORM_CONVERGENCE** | Pointwise convergence used as uniform | sup and limit swapped |\n| **MISSING_DOMINATION** | DCT cited but no dominating function given | |\n| **INTEGRABILITY_GAP** | Uses E|X|^p without proving\u002Fassuming finite moments | |\n| **REGULARITY_GAP** | Differentiability\u002FLipschitz\u002Fconvexity used but not established | |\n| **STOCHASTIC_MODE_CONFUSION** | Mixes a.s.\u002Fin prob.\u002Fin L²\u002Fin expectation | |\n\n### Group C: Model & Parameter Tracking\n\n| Category | Description | Example |\n|----------|-------------|---------|\n| **MISSING_DERIVATION** | A quantity is used but never derived from the model | Risk functional with undefined B, W |\n| **HIDDEN_ASSUMPTION** | Proof silently uses a condition not in the theorem | Gaussianity assumed but not stated |\n| **INSUFFICIENT_ASSUMPTION** | Hypotheses too weak for proof (counterexample exists) | Moment conditions admitting 2-point distributions |\n| **DIMENSION_TRACKING** | Parameter dependence (d, n, K, ...) not explicit | d enters only through κ |\n| **NORMALIZATION_MISMATCH** | Coordinate\u002Fscaling conventions inconsistent | Rescaled vs raw coordinates |\n| **CONSTANT_DEPENDENCE_HIDDEN** | \"C\" depends on d,n,K but treated as universal | |\n\n### Group D: Scope & Claims\n\n| Category | Description | Example |\n|----------|-------------|---------|\n| **SCOPE_OVERCLAIM** | Conclusion stated more broadly than proof supports | \"β_K=0\" with only generic overlap |\n| **REFERENCE_MISMATCH** | Cited theorem's hypotheses not verified at point of use | |\n\n## Two-Axis Severity System\n\n### Axis A — Proof Status (what is wrong)\n\n| Status | Meaning |\n|--------|---------|\n| **INVALID** | Statement false as written (counterexample exists or contradiction) |\n| **UNJUSTIFIED** | Could be true, but current proof does not establish it |\n| **UNDERSTATED** | True only after strengthening assumptions |\n| **OVERSTATED** | True only after weakening conclusion \u002F adding qualifiers |\n| **UNCLEAR** | Ambiguous notation \u002F definition drift (not wrong per se) |\n\n### Axis B — Impact (how much breaks)\n\n| Impact | Meaning |\n|--------|---------|\n| **GLOBAL** | Breaks main theorem or core dependency chain |\n| **LOCAL** | Affects a side result but not the main theorem |\n| **COSMETIC** | Exposition only |\n\n### Severity Labels (derived)\n\n| Label | Definition |\n|-------|------------|\n| **FATAL** | INVALID + GLOBAL |\n| **CRITICAL** | (INVALID + LOCAL) or (UNJUSTIFIED + GLOBAL) |\n| **MAJOR** | (UNJUSTIFIED + LOCAL) or (UNDERSTATED\u002FOVERSTATED + GLOBAL) |\n| **MINOR** | Clarity \u002F notation \u002F dimension bookkeeping that doesn't change claims |\n\n## Side-Condition Checklists for Common Theorems\n\nWhen the proof invokes any of the following, require explicit verification of ALL listed conditions:\n\n| Theorem | Required Conditions |\n|---------|-------------------|\n| **DCT** (Dominated Convergence) | Pointwise a.e. convergence + integrable dominating function |\n| **MCT** (Monotone Convergence) | Monotone increasing + non-negative |\n| **Fubini\u002FTonelli** | Product measurability + integrability (Fubini) or non-negative (Tonelli) |\n| **Leibniz integral rule** | Continuity of integrand + dominating function for derivative |\n| **Implicit Function Theorem** | Continuous differentiability + non-singular Jacobian |\n| **Taylor with remainder** | Sufficient differentiability + remainder form (Lagrange\u002Fintegral) |\n| **Jensen's inequality** | Convexity of function + integrability |\n| **Cauchy-Schwarz** | Correct inner product space + integrability of both factors |\n| **Weyl\u002FDavis-Kahan** | Symmetry\u002FHermiticity + perturbation bound conditions |\n| **Analytic continuation** | Domain connectivity + identity theorem conditions |\n| **WLOG reduction** | Invariance under claimed symmetry + reduction is reversible |\n\n## Workflow\n\n### Proof-obligation fan-out\n\nIndependent sections\u002Ftheorems may be extracted by fresh read-only\n`spawn_agent` shards, with a sequential fresh-context fallback. Each shard\nreturns `{\"shard_id\": ..., \"entries\": [{\"payload\": ..., \"dedup_key\":\n\"\u003Ctheorem-or-obligation-id>\"}]}` and must not declare the proof valid. The\nparent mechanically merges obligations; the fresh Codex review that evaluates\nthem records `review_independence: same-family` and\n`acceptance_status: provisional`. See\n[`fan-out-pattern.md`](..\u002Fshared-references\u002Ffan-out-pattern.md).\n\n### Phase 0: Preparation\n\n1. **Locate the proof**: Find the main `.tex` file(s).\n2. **Read the entire proof**: Extract list of all theorems\u002Flemmas\u002Fpropositions\u002Fcorollaries\u002Fdefinitions\u002Fassumptions.\n3. **Read reference materials**: Reference papers, prior results.\n4. **Build a section map**: Structured list with line numbers and key claims.\n5. **Identify the main theorem**: Central result, assumptions, claims.\n\n### Phase 0.5: Proof-Obligation Ledger\n\nBuild formal accounting artifacts. Save to `PROOF_SKELETON.md`:\n\n#### 1. Dependency DAG\nNodes = Definitions \u002F Assumptions \u002F Lemmas \u002F Theorems. Edges = \"uses\". **Detect cycles** (including semantic circularity where Lemma A uses a corollary that quietly depends on A).\n\n#### 2. Assumption Ledger\nFor each theorem\u002Flemma, list every hypothesis with WHERE each is verified (or mark \"UNVERIFIED\"). Track **usage-minimal assumption sets** — which assumptions were actually used vs merely stated.\n\n#### 3. Typed Symbol Table\nEach symbol must have a **type signature**:\n```\nκ : scalar ∈ (0,1), depends on (d, α_t, Σ, μ)\nu* : vector ∈ ℝ^d, u* = C^{-1}m\nB^even : matrix ∈ ℝ^{(L+1)×(L+1)}, symmetric PSD\nΨ_v : function ℝ → ℝ, analytic in (ζ,κ), parity determined by v\n```\nFlag any symbol whose meaning changes or whose type is inconsistent across uses.\n\n#### 4. Canonical Quantified Statements\nFor each theorem\u002Flemma, rewrite the statement with **explicit quantifiers, domains, and limit order**:\n```\n∀K ≥ 3, ∀π ∈ Π_K^{ms,∘} \\ E_K, ∃κ_0 > 0 such that ∀κ ∈ (0, κ_0):\n  h_act^{(K,π)} = Θ(κ^{α_K^act})  [uniform in π on compact subsets]\n```\nIf you cannot restate a theorem this precisely, mark it **UNCLEAR — needs disambiguation**.\n\n#### 5. Micro-Claim Inventory\nEvery nontrivial step becomes a numbered micro-claim in **sequent form**:\n```\nMC-17: Context: [Lemma 3.1, κ \u003C κ_0, Z_κ has bounded moments up to order 2m+2]\n       ⊢ Goal: P̂_0 is positive definite\n       Rule: monomials linearly independent on support of continuous distribution\n       Side-conditions: positive density near origin ✓ (by GMM weak convergence)\n```\nEach micro-claim has: justification rule name + required conditions + where conditions are proven.\n\n#### 6. Limit-Order Map\nTrack every asymptotic statement's **limit order and uniformity scope**:\n```\nh_act = Θ(κ^α)  [as κ→0, uniform in π on compact subsets of Π_K, for fixed K]\nτ_act ~ (b\u002Fa)n   [as n→∞, for fixed κ,K,π with x_K ≪ 1]\n```\nFlag any statement where limit order is ambiguous or uniformity is unclear.\n\n### Optional Lean verification\n\nWhen the user requests Lean or a specific obligation warrants formal checking,\ndelegate that obligation and its original hypotheses to\n[`\u002Flean-formalize`](..\u002Flean-formalize\u002FSKILL.md). Record the checked scope, exported\ndeclaration, statement-alignment evidence, build output and transitive-axiom\naudit, then resume this audit. A proved sublemma closes only the corresponding\nobligation. Keep the complete proof audit, acceptance decision and canonical wiki\nartifacts here; the nested Lean task returns evidence without starting another\n`proof-checker` run. Routine proof checks do not require Lean. Reviewer agreement\nand unsuccessful counterexample searches do not discharge a proof obligation.\n\n### Phase 1: First Review (Codex GPT-6-Astra ultra)\n\nSubmit the **complete proof content** with the following **mandatory reviewer checklist** in the prompt:\n\n```text\nspawn_agent:\n  model: gpt-6-astra\n  reasoning_effort: ultra\n  message: |\n    You are performing a rigorous mathematical proof review. For EVERY theorem,\n    lemma, and proposition, check ALL of the following:\n\n    ## MANDATORY CHECKS\n\n    A. DEFINITIONS: List any symbol whose meaning is ambiguous or changes.\n    B. HYPOTHESIS DISCHARGE: For each lemma\u002Ftheorem APPLICATION (not statement),\n       list each hypothesis and whether it was verified, with location.\n    C. INEQUALITY AUDIT: For each inequality chain, verify direction, missing\n       absolute values, missing conditions (convexity, PSD, integrability).\n    D. INTERCHANGE AUDIT: Flag every limit\u002Fderivative\u002Fexpectation\u002Fintegral\n       interchange. State which theorem justifies it (DCT\u002FMCT\u002FFubini\u002FLeibniz)\n       and which conditions are verified\u002Fmissing.\n    E. PROBABILITY MODE: Track whether claims are a.s.\u002Fin prob.\u002Fin expectation\u002F\n       w.h.p. Ensure transitions are justified.\n    F. UNIFORMITY & CONSTANTS: For every O(·), o(·), Θ(·), ≲, state whether\n       it is uniform over all parameters. List hidden parameter dependence.\n    G. EDGE\u002FDEGENERATE CASES: Attempt to break each key lemma with a 1D,\n       low-rank, or extreme-parameter construction.\n    H. DEPENDENCY CONSISTENCY: Detect cycles or forward references to unproven\n       results.\n\n    ## OUTPUT FORMAT (per issue)\n    For each issue found, provide:\n    - id: sequential number\n    - status: INVALID \u002F UNJUSTIFIED \u002F UNDERSTATED \u002F OVERSTATED \u002F UNCLEAR\n    - impact: GLOBAL \u002F LOCAL \u002F COSMETIC\n    - category: [from taxonomy]\n    - location: section\u002Fequation\u002Fline\n    - statement: what the proof claims\n    - why_invalid: why this is wrong or unjustified\n    - counterexample: YES (describe) \u002F NO \u002F CANDIDATE (describe attempt)\n    - affects: which downstream results break if this is wrong\n    - minimal_fix: how to fix it\n\n    [FULL PROOF CONTENT HERE]\n```\n\n**Save the reviewer `agent_id`.** Parse into structured issue list. Write to `PROOF_AUDIT.md`.\n\n### Phase 1.5: Counterexample Red Team\n\nFor each CRITICAL or MAJOR issue, and for every key lemma that introduces:\n- a new inequality bound\n- an identifiability\u002Funiqueness claim\n- a curvature\u002FPSD\u002Fstrong convexity assertion\n- a uniform-in-parameter claim\n- a convergence mode upgrade (pointwise → uniform, in prob → w.h.p.)\n\nSystematically attempt to construct counterexamples using:\n\n| Strategy | Description |\n|----------|-------------|\n| **Dimensional collapse** | Set d=1 or 2, K=2, n small |\n| **Degeneracy** | Singular covariance, tiny weight, overlapping means, identical components |\n| **Extremal distributions** | Two-point ±a, bounded non-subGaussian, heavy tails |\n| **Adversarial parameter scaling** | Pick parameters making neglected terms dominate |\n| **Numeric falsification** | Translate lemma to a function, brute-force optimize over small domain |\n\n**Rule**: Label \"counterexample found\" ONLY if algebraically verified. Otherwise log as \"candidate counterexample — needs verification.\"\n\nRecord all attempts (successful or not) in `PROOF_AUDIT.md`.\n\n### Phase 2: Fix Implementation\n\nFor each issue, ordered by severity (FATAL → CRITICAL → MAJOR → MINOR):\n\n#### Step 2a: Choose fix strategy\nFor each issue, explicitly choose one of:\n- **ADD_DERIVATION**: Write missing proof steps\n- **STRENGTHEN_ASSUMPTION**: Add conditions to theorem statement\n- **WEAKEN_CLAIM**: Reduce conclusion scope\n- **ADD_REFERENCE**: Cite known result + verify its conditions apply\n\nLog this choice — it is a scope-changing decision when it alters theorem statements.\n\n#### Step 2b: Derive the fix mathematically\n- Complete mathematical derivation, not just a claim\n- If new proposition\u002Flemma needed, write in full theorem-proof style\n\n#### Step 2c: Implement in LaTeX\n- Edit the `.tex` file\n- Preserve existing `\\label` references where possible\n\n#### Step 2d: Record the fix\n```markdown\n### Fix N: [SHORT TITLE]\n**Issue**: [id] [CATEGORY] — [description]\n**Severity**: FATAL \u002F CRITICAL \u002F MAJOR \u002F MINOR\n**Status**: INVALID \u002F UNJUSTIFIED \u002F UNDERSTATED \u002F OVERSTATED\n**Impact**: GLOBAL \u002F LOCAL \u002F COSMETIC\n**Fix strategy**: ADD_DERIVATION \u002F STRENGTHEN_ASSUMPTION \u002F WEAKEN_CLAIM \u002F ADD_REFERENCE\n**Location**: Section X, Lines Y-Z\n\n**BEFORE**: [what the proof originally did]\n**WHY WRONG**: [mathematical problem, with counterexample if applicable]\n**AFTER**: [what the fix does]\n**KEY EQUATION**: [central new equation]\n**PROOF OBLIGATIONS ADDED**: [new conditions\u002Flemmas introduced]\n**DOWNSTREAM EFFECTS**: [which results now need re-checking]\n```\n\n#### Step 2e: Compile check\n```bash\npdflatex -interaction=nonstopmode \u003Cfile>.tex 2>&1 | grep -E \"Error|Warning|undefined\"\n```\n\n### Phase 3: Re-Review (Codex GPT-6-Astra ultra)\n\nLaunch a fresh reviewer agent for the next review round. Do not use `send_input` here; proof-checker keeps each round independent. Request the same mandatory checklist.\n\nCheck acceptance gate. If not met, repeat Phases 2-3 (up to MAX_REVIEW_ROUNDS).\n\n### Phase 3.5: Global Closure & Independent Verification\n\n#### Global closure checks\nAfter all fixes, verify the proof as a whole:\n- **Statement–conclusion match**: Does the proof end with EXACTLY what the theorem claims (quantifiers, constants, uniformity)?\n- **All obligations discharged**: Every node in the obligation DAG is proven or explicitly assumed (and the theorem statement includes it).\n- **Case analysis coverage**: Cases partition the domain AND include boundary\u002Fdegenerate cases.\n- **Induction correctness** (if applicable): Base case, inductive step, correct use of IH, induction measure strictly decreases.\n- **WLOG reductions**: Each \"without loss of generality\" spawns a micro-claim proving the reduction is lossless.\n- **No silent assumption strengthening**: Any fix that strengthened assumptions has propagated to the main theorem statement.\n\n#### Independent second review for FATAL\u002FCRITICAL fixes\nFor any fix that resolved a FATAL or CRITICAL issue, submit the **fixed section alone** (without showing the previous critique) to a **fresh Codex thread**:\n\n```text\nspawn_agent:\n  model: gpt-6-astra\n  reasoning_effort: ultra\n  message: |\n    Blind review of the following proof section. You have NOT seen any prior\n    review or discussion. Check every step for correctness, hidden assumptions,\n    illegal interchanges, and counterexamples.\n    [FIXED SECTION ONLY]\n```\n\nIf the blind reviewer finds new issues, re-enter Phase 2.\n\n#### Regression proof-audit\nAfter fixes, re-run:\n- DAG acyclicity check (no new cycles introduced)\n- Counterexample suite on all DOWNSTREAM lemmas of modified results\n- Assumption-delta report: what became stronger\u002Fweaker due to fixes?\n\n### Phase 3.9: Unrecoverable Proof Protocol\n\nIf acceptance gate is not met after MAX_REVIEW_ROUNDS, output a **Proof Unrecoverable Report**:\n1. Minimal set of blocking FATAL\u002FCRITICAL issues that could not be resolved\n2. Salvage options ranked: (a) weaken claim, (b) strengthen assumptions, (c) add missing lemmas, (d) restructure argument\n3. Which parts of the proof are likely still reusable\n4. Recommended next steps for the author\n\nDo NOT silently declare success. The report must be honest.\n\n### Phase 4: Audit Report Generation\n\nGenerate `proof_audit_report.tex` with:\n\n1. **Overview table**: All issues with two-axis severity, category, fix strategy, status\n2. **Before\u002FAfter logic chain**: Red (BEFORE) → Green (AFTER) comparison\n3. **For each fix**: original proof → why wrong → counterexample (if any) → complete derivation → remaining subtleties\n4. **Proof-obligation diff**: What was unverified before, what is verified now\n5. **Summary**: Now proven \u002F still assumed \u002F open problems\n6. **Colored boxes**: BEFORE (red), AFTER (green), WHY WRONG (orange), KEY INSIGHT (blue), WARNING (yellow)\n\nCompile: `pdflatex proof_audit_report.tex && pdflatex proof_audit_report.tex`\n\n### Phase 5: State Persistence\n\nWrite `PROOF_CHECK_STATE.json`:\n```json\n{\n  \"status\": \"completed\",\n  \"rounds\": 2,\n  \"review_agent_ids\": [\"...\"],\n  \"fatal_fixed\": 0,\n  \"critical_fixed\": 3,\n  \"major_fixed\": 2,\n  \"minor_fixed\": 1,\n  \"counterexamples_found\": 1,\n  \"counterexample_candidates\": 2,\n  \"acceptance_gate\": \"PASS\",\n  \"timestamp\": \"...\"\n}\n```\n\n### Phase 5.5: Research Wiki Claim Ledger (additive; only if a wiki is active)\n\nIf — and only if — a `research-wiki\u002F` exists, persist each top-level theorem\u002Fheadline\nas a **claim node** (the wiki's PROVE\u002FJUDGE ledger). This is the **birth point** for wiki\nclaim nodes. It is a **detect-only record, never a verdict**: it never changes the audit's\n`verdict`\u002F`reason_code`, never blocks, and is skipped when `verdict == NOT_APPLICABLE` or no\nwiki is found. The claim's `status` is the **PROOF axis only** ({drafted, unproven,\nsound-modulo-imports, verified, refuted, retracted}); empirical experiment support is a\nseparate axis carried by edges (`\u002Fresult-to-claim`), never written into `status`.\n\nResolve the helper via the Codex-side chain (skip cleanly if unavailable; the audit is\nalready complete):\n```\nARIS_REPO=\"${ARIS_REPO:-$(awk -F'\\t' '$1==\"repo_root\"{print $2; exit}' .aris\u002Finstalled-skills-codex.txt 2>\u002Fdev\u002Fnull)}\"\nWIKI_SCRIPT=\"\"\n[ -n \"$ARIS_REPO\" ] && [ -f \"$ARIS_REPO\u002Ftools\u002Fresearch_wiki.py\" ] && WIKI_SCRIPT=\"$ARIS_REPO\u002Ftools\u002Fresearch_wiki.py\"\n[ -z \"$WIKI_SCRIPT\" ] && [ -f tools\u002Fresearch_wiki.py ] && WIKI_SCRIPT=\"tools\u002Fresearch_wiki.py\"\n[ -z \"$WIKI_SCRIPT\" ] && [ -f ~\u002F.codex\u002Fskills\u002Fresearch-wiki\u002Fresearch_wiki.py ] && WIKI_SCRIPT=\"$HOME\u002F.codex\u002Fskills\u002Fresearch-wiki\u002Fresearch_wiki.py\"\n```\n\nIf `research-wiki\u002F` exists and `WIKI_SCRIPT` is available and `verdict != NOT_APPLICABLE`,\nfor each top-level theorem map the audit outcome to an honest status — `PASS`\u002Fall proofs\ncomplete → `verified`; closes modulo flagged imports → `sound-modulo-imports`; counterexample\nfound or statement judged false → `refuted`; open gap (UNJUSTIFIED, no counterexample) →\n`unproven` (never fake a gap as `refuted`\u002F`verified`) — then record it (idempotent):\n```\npython3 \"$WIKI_SCRIPT\" add_claim research-wiki\u002F --slug \"\u003Cstable-theorem-id>\" \\\n     --name \"\u003Ctheorem headline>\" --status \"\u003Cmapped status>\" \\\n     --provenance \"\u003Ctrace_path from PROOF_AUDIT.json>\" --statement \"\u003Ccanonical statement>\" \\\n     --scope \"\u003Cwhat it does NOT say; flagged imports>\" --update-on-exist\n```\n`add_claim` failure is non-fatal (warn and continue; the audit is unaffected).\n\n## Key Rules\n\n### Mathematical rigor\n- **Never accept a proof step on faith**. \"Clearly\" \u002F \"it follows\" \u002F \"by standard arguments\" are red flags — each must spawn a micro-claim.\n- **Hypothesis discharge**: Every time a lemma is APPLIED, verify EACH of its hypotheses at that point. Use the side-condition checklists above.\n- **Interchange discipline**: Every swap of limit\u002Fexpectation\u002Fderivative\u002Fintegral must cite a theorem (DCT\u002FMCT\u002FFubini\u002FLeibniz) and verify its conditions with explicit dominating function or integrability proof.\n- **Uniformity discipline**: Every O(·)\u002FΘ(·) must declare what parameters it is uniform over. \"O(1)\" that secretly depends on d,n,K is a CONSTANT_DEPENDENCE_HIDDEN issue.\n- **Quantifier discipline**: Check ∀\u002F∃ order. \"For sufficiently small κ\" must specify: does κ₀ depend on K? On π? On d?\n- **Counterexample-first**: Before trying to fix a gap, first try to break it.\n- **WLOG prohibition**: Every \"without loss of generality\" must have an explicit micro-claim proving the reduction. No free WLOGs.\n- **No silent assumption strengthening**: Any fix that adds conditions must propagate to the theorem statement.\n\n### Review-independence protocol\n- **Codex executor analyzes and implements; a fresh Codex reviewer provides adversarial review.** Base review remains same-family\u002Fprovisional.\n- **Codex reasoning always ultra** (deep-audit tier): never below `xhigh` — only the capability fallback in `reviewer-routing.md` may step down, and only on explicit capability errors.\n- **Send full content**: Don't summarize — send actual math for line-by-line checking.\n- **Fresh reviewer agents**: Save each returned `agent_id` for traceability, but launch a new `spawn_agent` for each review round. Do not use `send_input` across proof-checker rounds.\n\n### Fix quality\n- **Minimal fixes**: Fix exactly what's broken, nothing more.\n- **Full derivation**: Every fix includes complete mathematical argument.\n- **Explicit scope decisions**: Each fix is tagged ADD_DERIVATION \u002F STRENGTHEN_ASSUMPTION \u002F WEAKEN_CLAIM \u002F ADD_REFERENCE.\n- **Compile after each fix**: LaTeX must compile cleanly.\n\n### Scope honesty\n- **Don't overclaim**: If a fix makes a result conditional, say so.\n- **Separate \"proven\" from \"assumed\"**: The audit report has an explicit section for this.\n- **Log open problems**: Issues requiring future work are listed, not hidden.\n\n## Output Files\n\n| File | Content | When |\n|------|---------|------|\n| `PROOF_SKELETON.md` | Dependency DAG + assumption ledger + micro-claims | Phase 0.5 |\n| `PROOF_AUDIT.md` | Cumulative round-by-round audit log | Updated each round |\n| `PROOF_AUDIT.json` | Machine-readable submission verdict (see below) | Always emitted |\n| `proof_audit_report.tex\u002F.pdf` | Formal before\u002Fafter report | Phase 4 |\n| `PROOF_CHECK_STATE.json` | State for recovery | Phase 5 |\n| `PROOF_AUDIT.html` (+ `.review.json` sidecar) | Single-file HTML view auto-rendered via `\u002Frender-html \"PROOF_AUDIT.md\" --json \"PROOF_AUDIT.json\"`. **Non-blocking** — if `\u002Frender-html` fails the audit still counts as complete. | Workflow end (when `RENDER_HTML = true`, default) |\n\n## Submission Artifact Emission\n\nThis skill **always** writes `PROOF_AUDIT.json` at the paper directory\nroot (i.e. `paper\u002FPROOF_AUDIT.json` when invoked from `\u002Fpaper-writing`\nwith paper-dir `paper\u002F`; `\u003Cyour-paper-dir>\u002FPROOF_AUDIT.json` when invoked\nstandalone), regardless of caller or whether the paper contains theorems.\nA paper with no `\\begin{theorem}` \u002F `\\begin{lemma}` \u002F `\\begin{proof}` emits\nverdict `NOT_APPLICABLE`; silent skip is forbidden. `paper-writing`\nPhase 6 and `verify_paper_audits.sh` both rely on this artifact\nexisting at `\u003Cpaper-dir>\u002FPROOF_AUDIT.json`.\n\nThe artifact conforms to the schema in `shared-references\u002Fassurance-contract.md`:\n\n```json\n{\n  \"audit_skill\":      \"proof-checker\",\n  \"verdict\":          \"PASS | WARN | FAIL | NOT_APPLICABLE | BLOCKED | ERROR\",\n  \"reason_code\":      \"all_proofs_complete | minor_gaps | critical_gap | no_theorems | ...\",\n  \"summary\":          \"One-line human-readable verdict summary.\",\n  \"audited_input_hashes\": {\n    \"main.tex\":                 \"sha256:...\",\n    \"sections\u002F4.theory.tex\":    \"sha256:...\"\n  },\n  \"trace_path\":       \".aris\u002Ftraces\u002Fproof-checker\u002F\u003Cdate>_run\u003CNN>\u002F\",\n  \"thread_id\":        \"\u003Ccodex mcp thread id>\",\n  \"executor_model\":   \"codex-gpt-6-astra\",\n  \"executor_family\":  \"openai\",\n  \"reviewer_model\":   \"gpt-6-astra\",\n  \"reviewer_family\":  \"openai\",\n  \"review_independence\": \"same-family\",\n  \"acceptance_status\": \"provisional\",\n  \"reviewer_reasoning\": \"ultra\",\n  \"generated_at\":     \"\u003CUTC ISO-8601>\",\n  \"details\": {\n    \"theorems_audited\": \u003Cint>,\n    \"issues\": [ { \"id\": \"T1-H3\", \"severity\": \"FATAL|CRITICAL|MAJOR|MINOR\",\n                  \"category\": \"quantifier|domination|...\",\n                  \"location\": \"sections\u002F4.theory.tex:L182\",\n                  \"note\": \"...\" }, ... ]\n  }\n}\n```\n\n### `audited_input_hashes` scope\n\nHash the **declared input set** actually reviewed — the theorem-bearing\n`.tex` files passed into this invocation — not a repo-wide union and not\nthe reviewer's self-reported opened subset. The external verifier rehashes\nthese entries; any mismatch flags `STALE`.\n\n**Path convention** (must match `verify_paper_audits.sh`): keys are\n**paths relative to the paper directory** (no `paper\u002F` prefix — the\nverifier resolves relative to the paper dir; prefixing produces\n`paper\u002Fpaper\u002F...` and false-fails as STALE). Use **absolute paths** for\nfiles outside the paper dir.\n\n### Verdict decision table\n\n| Input state                                           | Verdict          | `reason_code` example |\n|-------------------------------------------------------|------------------|-----------------------|\n| No theorems \u002F lemmas \u002F proofs in paper                | `NOT_APPLICABLE` | `no_theorems`         |\n| Theorems present but referenced files unreadable      | `BLOCKED`        | `source_unreadable`   |\n| All proof obligations discharged, no gaps             | `PASS`           | `all_proofs_complete` |\n| Only MINOR issues (notation \u002F exposition)             | `WARN`           | `minor_gaps`          |\n| Any FATAL or CRITICAL issue (logic gap, wrong claim)  | `FAIL`           | `critical_gap`        |\n| Reviewer invocation failed (network \u002F malformed)      | `ERROR`          | `reviewer_error`      |\n\nMAJOR issues alone map to `WARN` or `FAIL` at the reviewer's discretion and\nmust carry an explicit justification in `summary` + `details.issues`.\n\n### Thread independence\n\nEvery invocation uses a fresh reviewer agent. Never use `send_input` across\nproof-checker runs. Do not accept prior audit outputs\n(PAPER_CLAIM_AUDIT, CITATION_AUDIT, EXPERIMENT_LOG) as input — the fresh\nthread preserves reviewer independence per\n`shared-references\u002Freviewer-independence.md`.\n\nThis skill never blocks by itself; `paper-writing` Phase 6 plus the\nverifier decide whether the verdict blocks finalization based on the\n`assurance` level.\n\n## Example Invocations\n\n```\n\u002Fproof-checker \"neurips_2025.tex\"\n\u002Fproof-checker \"check the GMM generalization proof, focus on dimension dependence\"\n\u002Fproof-checker \"verify proof in paper.tex — difficulty: nightmare\"\n```\n",{"name":5,"description":7,"allowed-tools":24,"argument-hint":25},"Bash(*), Read, Grep, Glob, Write, Edit","[path-to-tex-file or proof-description]","main","18193d71ea3441d7c033dbf195b8a7c75ac96f75","packages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fskills-codex\u002Fproof-checker",0,"https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science","https:\u002F\u002Fgithub.com\u002FCZX-LBJ","一个没事喜欢随便写点代码的小菜鸡\r\n","2026-10-22 09:48:04",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add CZX-LBJ\u002Fdeepseek-science --skill proof-checker",{"tool":40,"command":41},"bunx","bunx skills add CZX-LBJ\u002Fdeepseek-science --skill proof-checker",{"tool":43,"command":44},"pnpm","pnpm dlx skills add CZX-LBJ\u002Fdeepseek-science --skill proof-checker","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fskills-codex\u002Fproof-checker\u002FSKILL.md",true,[],[49,56,63,70,77,84],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":19},"01M7FEEKSNMRMKW6JEK93PFDFY","bio-free-energy-calculations","Performs alchemical free-energy calculations including relative binding free energy (RBFE \u002F FEP+) and absolute binding free energy (ABFE) via OpenFE, FEP+, GROMACS, AMBER pmemd, and OpenMM with explicit lambda scheduling, soft-core potentials, MBAR\u002FBAR analysis, cycle-closure validation, and protocol-appropriate enhanced sampling. Compares ML alternatives (Boltz-2 affinity, DeepDock). Use when ranking analogs by binding affinity beyond docking accuracy, performing prospective lead optimization, or validating SAR predictions.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Ffree-energy-calculations\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M7FEE5CTE5QK55CEMXNNVWY9","bio-conformer-generation","Generates 3D conformer ensembles using RDKit ETKDGv3 with knowledge-enhanced distance geometry, MMFF94\u002FUFF force-field optimization, CREST + GFN2-xTB semi-empirical refinement, and macrocycle-aware torsion preferences. Provides explicit decision rules for single vs ensemble conformer use, RMSD pruning, energy windows, conformer count, and force-field choice. Use when preparing 3D ligands for docking, generating descriptor input for 3D QSAR, or sampling macrocycle\u002Fpeptide conformational ensembles.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Fconformer-generation\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M7FEDNGFEMFWD1MSEDR5H7CY","bio-admet-prediction","Predicts ADMET properties using ADMETlab 3.0 (119 platform features, including 77 prediction models with modeled-endpoint uncertainty), ADMET-AI, DeepChem MolNet, and chemprop D-MPNN with explicit handling of OECD QSAR principles, applicability domain assessment, calibration, hERG\u002FCYP\u002FAMES endpoints, and PAINS \u002F Lipinski \u002F Ro5 \u002F Veber \u002F BBB druglikeness filters. Use when filtering compounds for drug-likeness, prioritizing leads by predicted safety, or building an in-house ADMET QSAR model.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Fadmet-prediction\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7FED69GFB54FRF7AV5066DN","writing-systems-papers","Paragraph-level structural blueprint for 10-12 page systems papers targeting OSDI, SOSP, ASPLOS, NSDI, and EuroSys. Provides page allocation, paragraph templates, and writing patterns. Use when user says \"写系统论文\", \"systems paper structure\", \"OSDI paper\", \"SOSP paper\", or wants fine-grained structural guidance for a systems conference submission.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fwriting-systems-papers\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7FECVVDWZ0X0VEDTCE4B4DZ","wiki-enrich","Fill in the per-paper TODO sections of research-wiki\u002Fpapers\u002F\u003Cslug>.md pages that literature-ingest skills leave as bare scaffolds. Use when user says 'enrich wiki', 'fill paper TODOs', 'wiki body 補完', '把 paper 摘要寫進 wiki', 'research-wiki 自動填', or after a batch ingest that left papers\u002F as TODO scaffolds.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fwiki-enrich\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M7FECD4BCJ2MTSBDDSE3ZAJ5","web-debug-search","Search GitHub, Stack Exchange, Chinese technical communities, official documentation, and general developer web sources for software errors, compatibility problems, API usage questions, and real-world workarounds. Use for debugging and discovery only; results are not paper-citation evidence.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fweb-debug-search\u002FSKILL.md",[],[]]