[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6CF23WT5GKGCKP1CVPXTHTW":3,"skill-same:01M6CF23WT5GKGCKP1CVPXTHTW":85},{"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":31,"publishedAt":46,"linkOnly":47,"markdownContent":48,"markdownBody":49,"frontMatter":50,"refName":63,"commitSha":64,"skillDirectory":65,"forksCount":66,"repositoryHtmlUrl":67,"repositoryHomepageUrl":6,"ownerHtmlUrl":68,"ownerBlogUrl":6,"ownerDescription":69,"updatedAt":70,"install":71,"resources":84},"01M6CF23WT5GKGCKP1CVPXTHTW","beta",null,"β role in CDR. Audits α's research matter for claim\u002Fevidence alignment, falsifiability, reproduction, citation integrity, and data-policy compliance.","usurobor\u002Fcnos","usurobor","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F182342826?v=4","01M6CEXJCRMYD8HB2315FGB1Q8","Peter Lisovin","01M6CEXJH2RW04GQ2SNXY40F5S",9,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos\u002Fblob\u002Fmain\u002Fsrc\u002Fpackages\u002Fcnos.cdr\u002Fskills\u002Fcdr\u002Fbeta\u002FSKILL.md",[18,23,28],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000021","资料检索与研究","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":26,"taxonomyTypeDesc":27},"01JBSEEDCAT000000000000033","数据分析师","PROFESSION","适用职业",{"categoryNo":29,"categoryName":30,"taxonomyType":26,"taxonomyTypeDesc":27},"01JBSEEDCAT000000000000037","律师与法务",[5,32,33,34,35,36,37,38,39,40,41,42,43,44,45],"policy","research","alpha","audit","compliance","data","SET","CDR","integrity","constraints","how","evidence","embedded","alignment","2026-11-10 10:12:30",false,"---\nname: beta\ndescription: β role in CDR. Audits α's research matter for claim\u002Fevidence alignment, falsifiability, reproduction, citation integrity, and data-policy compliance.\nartifact_class: skill\ngoverning_question: How does β preserve independent judgment over the research matter and reject claims stronger than the evidence supports?\nparent: cdr\ntriggers:\n  - beta\nscope: role-local\nkata_surface: embedded\ninputs:\n  - alpha research matter\n  - the claim and evidence set\n  - data-policy constraints\noutputs:\n  - independent beta audit verdict on claim and evidence alignment\n  - falsifiability, reproduction, and citation-integrity findings\n---\n\n# Beta (research β)\n\n> **This is a CDR-specific extension of the generic cnos.cdd β doctrine.** The\n> kernel grammar (role-cell shape, β-as-reviewer position, independence\n> requirement, review-then-close-out structure) is inherited by reference\n> from [`cnos.cdd\u002Fskills\u002Fcdd\u002Fbeta\u002FSKILL.md`](..\u002F..\u002F..\u002F..\u002Fcnos.cdd\u002Fskills\u002Fcdd\u002Fbeta\u002FSKILL.md).\n> Only the **review oracle** (the criterion β uses to evaluate matter) and\n> the **close-out** (the typed `#CDRReceipt` rather than a `git merge`)\n> diverge for the research loss function per\n> [`ROLES.md §4a.2`](..\u002F..\u002F..\u002F..\u002F..\u002F..\u002Fdocs\u002Fconcepts\u002FROLES.md#4a2-loss-function-distinction)\n> and the binding doctrinal contract [`CDR.md`](..\u002FCDR.md) Field 2 + Field 3.\n\n## Core Principle\n\n**Coherent research β work preserves independent judgment over the research matter, runs each Field-2 review oracle mechanically, and refuses to certify a claim whose calibration exceeds its evidence.**\n\nResearch β owns:\n- the review oracles per `CDR.md §\"Field 2\"` (falsifiability, diagnostic oracles, reproduction-from-clean, citation integrity, data-policy compliance, claim\u002Fevidence alignment)\n- the audit verdict recorded in the candidate `#CDRReceipt`'s gate-verdict surface (GO \u002F REVISE \u002F NO-GO \u002F INDETERMINATE \u002F BOUNDED-GO per `CDR.md §\"Field 3\"`)\n- the β-side close-out (review-context narrative); the typed receipt itself is γ's close-out artifact, not β's\n\nResearch β does **not** own a `git merge` analogue. Research has no engineering-style \"ship under repairable feedback\" surface; the matter that closes a wave is the typed receipt, validated by V (`cnos.cdd`'s validator with `protocol_id` dispatch per `schemas\u002Fcdd\u002FREADME.md`). β's verdict is recorded in the receipt; γ closes the wave.\n\nThe failure mode is **overclaim certification**: β certifies a claim whose evidence does not support its calibration; β accepts a receipt missing a reproduction record where one was required; β passes a citation that does not actually support the claim it backs. Per `CDR.md §0 Purpose`, the structural mechanism by which CDR resists overclaim is **β's independence from α**; collapsing β onto α within a single research-claim cycle is prohibited unconditionally (Field 6).\n\n## Load Order\n\nWhen acting as research β:\n\n1. Load [`CDR.md`](..\u002FCDR.md) — Field 2 (review oracles); Field 3 (close-out + verdict vocabulary); Field 6 (actor-collapse constraint; α=β prohibition for research-claim cycles).\n2. Load this file as the research-β role surface.\n3. Load the generic [`cnos.cdd\u002Fskills\u002Fcdd\u002Fbeta\u002FSKILL.md`](..\u002F..\u002F..\u002F..\u002Fcnos.cdd\u002Fskills\u002Fcdd\u002Fbeta\u002FSKILL.md) for the kernel-grammar reference — independence rule, \"doc is hypothesis; code is evidence\" (anchor oracle evidence on the artifact, not on auxiliary documentation), refusal of operator-direct instructions during the cycle. Engineering-specific subsections (pre-merge gate row 3's CI-equivalent enumeration; release\u002FSKILL.md) do not transfer; their research analogues are named in §\"Pre-receipt-acceptance gate\" below.\n4. Load the typed receipt schema [`schemas\u002Fcdr\u002Freceipt.cue`](..\u002F..\u002F..\u002F..\u002F..\u002F..\u002Fschemas\u002Fcdr\u002Freceipt.cue) — `#CDRReceipt` is the surface β audits against.\n5. Load Tier 2 and Tier 3 skills as required by the matter (statistical-method audit skills, citation-resolution skills, dataset-provenance audit skills as the project binding declares them).\n\n## Review oracles (per `CDR.md §\"Field 2\"`)\n\nResearch β runs each of the following oracles on every wave; each oracle is mechanically checkable or procedurally enforceable. The primary failure each catches is **overclaim** — a claim made stronger than its evidence supports.\n\n### 1. Falsifiability\n\nA claim is reviewable only if some specifiable observation would falsify it. β rejects claims too vague to falsify (e.g. \"the system behaves well\"). A falsifiable restatement is required (e.g. \"the measurement decreases monotonically under condition X\"). Vacuous claims drive a REVISE verdict.\n\n### 2. Diagnostic oracles\n\nWhere claims rest on measurements, the measurement procedure has a known failure-detection step: calibration check, signal-to-noise floor, control comparison, sentinel test. β verifies the diagnostic ran and that its result is recorded in the receipt. A claim that asserts a measurement without naming the diagnostic that would catch the measurement failing is incomplete matter.\n\n### 3. Reproduction-from-clean\n\nFor `claim_status ∈ {observed, computed}`, β re-runs the producing command from a clean environment per `CDR.md §\"Field 2\"`. The recorded output must match what the receipt claims. The reproduction record carries the canonical command (with args + SHA) and an output-match boolean. β does not skip this oracle for \"obvious\" claims — the reproduction surface is the structural mechanism by which research-claim transmission is bounded.\n\n### 4. Citation integrity\n\nEvery claim derived from external work cites a resolvable reference; every cited reference supports the claim it is invoked for. β rejects receipts whose citations do not resolve or whose invocation overstates the cited result (e.g. citing a paper that showed effect X to support a stronger claim Y). The check is *code-first* (or in the research case, *receipt-first*): re-read the cited source against the citing claim; do not anchor on the doc's summary alone (this is the research analogue of the engineering β's \"anchor oracle evidence on code, not doc\" rule).\n\n### 5. Data-policy compliance\n\n`data_refs` comply with the project's data-use policy (consent, anonymisation, retention, redistribution rights). β rejects receipts whose data refs violate policy or whose data refs are under-specified (no manifest, no checksum). The project binding (`\u003Cproject>\u002F.cdr\u002FPOLICY.md` or equivalent) defines the specifics; β verifies compliance.\n\n### 6. Claim\u002Fevidence alignment\n\nThe claim's strength is no stronger than the evidence supports. Verified against α's pre-emission table:\n\n| claim_status | required evidence refs | required reproduction |\n|---|---|---|\n| `observed` | `data_refs` (≥1, with manifest + checksum) | required |\n| `computed` | `method_refs` (≥1, script + SHA) + `data_refs` + `result_refs` | required |\n| `inferred` | inferential step named; supporting claims cited via `claim_refs` | not required for the inference itself |\n| `hypothesized` | none required | not required |\n| `indeterminate` | what would change the determination is recorded | not required |\n\nA claim whose calibration exceeds the evidence is the canonical overclaim. β returns REVISE with the calibration weakening (or evidence strengthening) named.\n\n## Engineering's \"compiles + tests pass\" does not apply\n\n`CDR.md §\"Field 2\"` declares: \"Engineering's 'compiles + tests pass' oracle does **not** apply. Research's truth is not provable by execution alone; the oracle is evidential alignment, not execution success.\" Research β does not run `go test` or its analogues against the matter as a primary oracle. Test execution may be relevant when the matter includes a method (a script that produces evidence) — in which case the method's execution and the reproduction record converge — but execution alone never certifies a research claim.\n\n## Verdict vocabulary (per `CDR.md §\"Field 3\"`)\n\nResearch β records the verdict in the candidate `#CDRReceipt`'s gate surface:\n\n- **GO** — the claim is transmissible at full strength. Maps to `boundary_decision.action: accept` + `transmissibility: accepted`.\n- **REVISE** — the claim is transmissible after named revisions. Maps to `boundary_decision.action: repair_dispatch`. β returns the receipt for revision with the calibration \u002F evidence \u002F citation gap named.\n- **NO-GO** — the claim is not transmissible. Maps to `boundary_decision.action: reject`.\n- **INDETERMINATE** — evidence is insufficient to determine transmissibility. The receipt carries `claim_status: indeterminate` for the affected claim(s); the wave records what would change the determination.\n- **BOUNDED-GO** — the claim is transmissible within named bounds. Maps to `transmissibility: degraded` with `limitations` enumerating the bound.\n\nβ does **not** \"approve with follow-up.\" A claim that requires a known revision before transmissibility gets REVISE; a known caveat that is acceptable within scope gets BOUNDED-GO with the caveat in `limitations`. No undeclared deferrals.\n\n## Pre-receipt-acceptance gate\n\nBefore β records GO or BOUNDED-GO in the receipt (the research analogue of the engineering β's pre-merge gate), β verifies:\n\n1. **Identity truth** — β's commit-time git identity matches `beta@cdr.\u003Cproject>` (or its elision form per `cnos.cdd\u002Fskills\u002Fcdd\u002Foperator\u002FSKILL.md` §\"Git identity for role actors\"). The same identity-truth row as engineering β; research has no exemption.\n2. **Candidate receipt parses against `schemas\u002Fcdr\u002Freceipt.cue`** — `#CDRReceipt` validation passes; `protocol_id: cnos.cdd.cdr.receipt.v1`.\n3. **Each Field-2 oracle ran** — falsifiability, diagnostic oracles, reproduction-from-clean (for observed\u002Fcomputed), citation integrity, data-policy compliance, claim\u002Fevidence alignment. The β-side audit record names which oracle ran and with what result.\n4. **Reproduction-from-clean executed on β's environment** — not on α's. β cannot certify reproduction by reading α's reproduction record; β re-runs the canonical command in a clean environment β controls. Output match recorded.\n5. **Citations re-resolved β-side** — every citation re-read from the source; no anchoring on α's summary of the cited work.\n6. **Limitations sufficient** — the limitations the receipt declares cover the epistemic gaps β observed; no surfaced gap is left undeclared.\n\nRow 3 of the engineering β gate (validators + CI-equivalent on the merge tree) is replaced by row 3 + row 4 here: the validator is V (`cn-cdd-verify` dispatching by `protocol_id`); the CI-equivalent is the reproduction-from-clean. Engineering's R5-activate kata equivalent and frontmatter validation do not transfer mechanically.\n\n## Independence rules (research-specific)\n\n### 1. α=β collapse is never permitted for research-claim cycles\n\nPer `CDR.md §\"Field 6\"`, \"Always prohibited. Research β reviews α's claim\u002Fevidence alignment for overclaim; α reviewing α's own claims is order-0 observation masquerading as order-1.\" No research-class waiver exists. Engineering-class collapse precedents (γ+α+β-collapsed-on-δ for mechanical refactor cycles) do **not** transfer to research-class claim transmission — overclaim is precisely the mode α cannot self-detect.\n\nA wave whose receipt cannot distinguish α and β authorship (single signature, single session, no review record) is structurally invalid for research-class claim transmission. The project binding (or δ as fallback) must reject the receipt and re-run the wave with separated actors.\n\n### 2. β refuses operator-direct instructions during a wave\n\nDuring an active research wave, β communicates only with γ via the artifact channel (the wave's `.cdr\u002Fwaves\u002F{wave-id}\u002F` directory or equivalent project-binding surface). β does not take review instructions directly from the operator — γ is the interface for all coordination during the wave. This is the engineering β's \"refuse operator-direct instructions\" rule transposed to research.\n\n### 3. β authors its verdict; β does not author α's fix\n\nIf β returns REVISE, α performs the fix (re-calibrate a claim, add a missing data ref, re-run reproduction, strengthen a citation, weaken a limitation). β does not patch α's matter. Independence is destroyed if β authors the fix it is judging.\n\n### 4. Doctrinal-vocabulary-first; anchor on the typed receipt\n\nWhen auditing claim\u002Fevidence alignment, β re-reads the typed `#CDRReceipt` and the cited evidence sources *first*, then verifies that any narrative report or wave summary matches what the typed receipt actually emits. Narrative-first auditing produces an honest-looking pass when the narrative has drifted from the receipt. The narrative is hypothesis; the typed receipt + cited evidence are the authoritative surfaces. This is the research analogue of the engineering β's \"Anchor oracle evidence on code, not doc\" rule.\n\n## Persona \u002F protocol \u002F project boundary\n\nThis overlay declares **what research β does at the protocol layer**. It does not declare:\n\n- **Who is doing the review** (persona — layer 1). The persona hub's `## Discipline` section names the discipline profile (primary virtue, primary error). The protocol overlay names the role function; the persona hub names the persona's loss function.\n- **What concrete data policy applies** (project — layer 4). The data-policy oracle (§\"Review oracles\" #5) cites the project binding's `POLICY.md`; this overlay does not specify retention windows, consent forms, or anonymisation thresholds. Projects bind those.\n\n## Resumption\n\nThe generic β doctrine's resumption protocol applies. Research β's resumption cases:\n\n- **Wave β-review file** (`.cdr\u002Fwaves\u002F{wave-id}\u002Fbeta-review.md` or equivalent) — round appendices when β resumes after session interruption mid-audit. Typical sections: `[Verdict, Oracle Results, Reproduction Record, Citation Integrity, Findings]`.\n- **Multi-round audits** — each round is an append operation, never restart completed rounds.\n\n**Never restart completed sections.** Committed sections represent settled β judgment; resumption preserves that judgment and continues forward.\n","# Beta (research β)\n\n> **This is a CDR-specific extension of the generic cnos.cdd β doctrine.** The\n> kernel grammar (role-cell shape, β-as-reviewer position, independence\n> requirement, review-then-close-out structure) is inherited by reference\n> from [`cnos.cdd\u002Fskills\u002Fcdd\u002Fbeta\u002FSKILL.md`](..\u002F..\u002F..\u002F..\u002Fcnos.cdd\u002Fskills\u002Fcdd\u002Fbeta\u002FSKILL.md).\n> Only the **review oracle** (the criterion β uses to evaluate matter) and\n> the **close-out** (the typed `#CDRReceipt` rather than a `git merge`)\n> diverge for the research loss function per\n> [`ROLES.md §4a.2`](..\u002F..\u002F..\u002F..\u002F..\u002F..\u002Fdocs\u002Fconcepts\u002FROLES.md#4a2-loss-function-distinction)\n> and the binding doctrinal contract [`CDR.md`](..\u002FCDR.md) Field 2 + Field 3.\n\n## Core Principle\n\n**Coherent research β work preserves independent judgment over the research matter, runs each Field-2 review oracle mechanically, and refuses to certify a claim whose calibration exceeds its evidence.**\n\nResearch β owns:\n- the review oracles per `CDR.md §\"Field 2\"` (falsifiability, diagnostic oracles, reproduction-from-clean, citation integrity, data-policy compliance, claim\u002Fevidence alignment)\n- the audit verdict recorded in the candidate `#CDRReceipt`'s gate-verdict surface (GO \u002F REVISE \u002F NO-GO \u002F INDETERMINATE \u002F BOUNDED-GO per `CDR.md §\"Field 3\"`)\n- the β-side close-out (review-context narrative); the typed receipt itself is γ's close-out artifact, not β's\n\nResearch β does **not** own a `git merge` analogue. Research has no engineering-style \"ship under repairable feedback\" surface; the matter that closes a wave is the typed receipt, validated by V (`cnos.cdd`'s validator with `protocol_id` dispatch per `schemas\u002Fcdd\u002FREADME.md`). β's verdict is recorded in the receipt; γ closes the wave.\n\nThe failure mode is **overclaim certification**: β certifies a claim whose evidence does not support its calibration; β accepts a receipt missing a reproduction record where one was required; β passes a citation that does not actually support the claim it backs. Per `CDR.md §0 Purpose`, the structural mechanism by which CDR resists overclaim is **β's independence from α**; collapsing β onto α within a single research-claim cycle is prohibited unconditionally (Field 6).\n\n## Load Order\n\nWhen acting as research β:\n\n1. Load [`CDR.md`](..\u002FCDR.md) — Field 2 (review oracles); Field 3 (close-out + verdict vocabulary); Field 6 (actor-collapse constraint; α=β prohibition for research-claim cycles).\n2. Load this file as the research-β role surface.\n3. Load the generic [`cnos.cdd\u002Fskills\u002Fcdd\u002Fbeta\u002FSKILL.md`](..\u002F..\u002F..\u002F..\u002Fcnos.cdd\u002Fskills\u002Fcdd\u002Fbeta\u002FSKILL.md) for the kernel-grammar reference — independence rule, \"doc is hypothesis; code is evidence\" (anchor oracle evidence on the artifact, not on auxiliary documentation), refusal of operator-direct instructions during the cycle. Engineering-specific subsections (pre-merge gate row 3's CI-equivalent enumeration; release\u002FSKILL.md) do not transfer; their research analogues are named in §\"Pre-receipt-acceptance gate\" below.\n4. Load the typed receipt schema [`schemas\u002Fcdr\u002Freceipt.cue`](..\u002F..\u002F..\u002F..\u002F..\u002F..\u002Fschemas\u002Fcdr\u002Freceipt.cue) — `#CDRReceipt` is the surface β audits against.\n5. Load Tier 2 and Tier 3 skills as required by the matter (statistical-method audit skills, citation-resolution skills, dataset-provenance audit skills as the project binding declares them).\n\n## Review oracles (per `CDR.md §\"Field 2\"`)\n\nResearch β runs each of the following oracles on every wave; each oracle is mechanically checkable or procedurally enforceable. The primary failure each catches is **overclaim** — a claim made stronger than its evidence supports.\n\n### 1. Falsifiability\n\nA claim is reviewable only if some specifiable observation would falsify it. β rejects claims too vague to falsify (e.g. \"the system behaves well\"). A falsifiable restatement is required (e.g. \"the measurement decreases monotonically under condition X\"). Vacuous claims drive a REVISE verdict.\n\n### 2. Diagnostic oracles\n\nWhere claims rest on measurements, the measurement procedure has a known failure-detection step: calibration check, signal-to-noise floor, control comparison, sentinel test. β verifies the diagnostic ran and that its result is recorded in the receipt. A claim that asserts a measurement without naming the diagnostic that would catch the measurement failing is incomplete matter.\n\n### 3. Reproduction-from-clean\n\nFor `claim_status ∈ {observed, computed}`, β re-runs the producing command from a clean environment per `CDR.md §\"Field 2\"`. The recorded output must match what the receipt claims. The reproduction record carries the canonical command (with args + SHA) and an output-match boolean. β does not skip this oracle for \"obvious\" claims — the reproduction surface is the structural mechanism by which research-claim transmission is bounded.\n\n### 4. Citation integrity\n\nEvery claim derived from external work cites a resolvable reference; every cited reference supports the claim it is invoked for. β rejects receipts whose citations do not resolve or whose invocation overstates the cited result (e.g. citing a paper that showed effect X to support a stronger claim Y). The check is *code-first* (or in the research case, *receipt-first*): re-read the cited source against the citing claim; do not anchor on the doc's summary alone (this is the research analogue of the engineering β's \"anchor oracle evidence on code, not doc\" rule).\n\n### 5. Data-policy compliance\n\n`data_refs` comply with the project's data-use policy (consent, anonymisation, retention, redistribution rights). β rejects receipts whose data refs violate policy or whose data refs are under-specified (no manifest, no checksum). The project binding (`\u003Cproject>\u002F.cdr\u002FPOLICY.md` or equivalent) defines the specifics; β verifies compliance.\n\n### 6. Claim\u002Fevidence alignment\n\nThe claim's strength is no stronger than the evidence supports. Verified against α's pre-emission table:\n\n| claim_status | required evidence refs | required reproduction |\n|---|---|---|\n| `observed` | `data_refs` (≥1, with manifest + checksum) | required |\n| `computed` | `method_refs` (≥1, script + SHA) + `data_refs` + `result_refs` | required |\n| `inferred` | inferential step named; supporting claims cited via `claim_refs` | not required for the inference itself |\n| `hypothesized` | none required | not required |\n| `indeterminate` | what would change the determination is recorded | not required |\n\nA claim whose calibration exceeds the evidence is the canonical overclaim. β returns REVISE with the calibration weakening (or evidence strengthening) named.\n\n## Engineering's \"compiles + tests pass\" does not apply\n\n`CDR.md §\"Field 2\"` declares: \"Engineering's 'compiles + tests pass' oracle does **not** apply. Research's truth is not provable by execution alone; the oracle is evidential alignment, not execution success.\" Research β does not run `go test` or its analogues against the matter as a primary oracle. Test execution may be relevant when the matter includes a method (a script that produces evidence) — in which case the method's execution and the reproduction record converge — but execution alone never certifies a research claim.\n\n## Verdict vocabulary (per `CDR.md §\"Field 3\"`)\n\nResearch β records the verdict in the candidate `#CDRReceipt`'s gate surface:\n\n- **GO** — the claim is transmissible at full strength. Maps to `boundary_decision.action: accept` + `transmissibility: accepted`.\n- **REVISE** — the claim is transmissible after named revisions. Maps to `boundary_decision.action: repair_dispatch`. β returns the receipt for revision with the calibration \u002F evidence \u002F citation gap named.\n- **NO-GO** — the claim is not transmissible. Maps to `boundary_decision.action: reject`.\n- **INDETERMINATE** — evidence is insufficient to determine transmissibility. The receipt carries `claim_status: indeterminate` for the affected claim(s); the wave records what would change the determination.\n- **BOUNDED-GO** — the claim is transmissible within named bounds. Maps to `transmissibility: degraded` with `limitations` enumerating the bound.\n\nβ does **not** \"approve with follow-up.\" A claim that requires a known revision before transmissibility gets REVISE; a known caveat that is acceptable within scope gets BOUNDED-GO with the caveat in `limitations`. No undeclared deferrals.\n\n## Pre-receipt-acceptance gate\n\nBefore β records GO or BOUNDED-GO in the receipt (the research analogue of the engineering β's pre-merge gate), β verifies:\n\n1. **Identity truth** — β's commit-time git identity matches `beta@cdr.\u003Cproject>` (or its elision form per `cnos.cdd\u002Fskills\u002Fcdd\u002Foperator\u002FSKILL.md` §\"Git identity for role actors\"). The same identity-truth row as engineering β; research has no exemption.\n2. **Candidate receipt parses against `schemas\u002Fcdr\u002Freceipt.cue`** — `#CDRReceipt` validation passes; `protocol_id: cnos.cdd.cdr.receipt.v1`.\n3. **Each Field-2 oracle ran** — falsifiability, diagnostic oracles, reproduction-from-clean (for observed\u002Fcomputed), citation integrity, data-policy compliance, claim\u002Fevidence alignment. The β-side audit record names which oracle ran and with what result.\n4. **Reproduction-from-clean executed on β's environment** — not on α's. β cannot certify reproduction by reading α's reproduction record; β re-runs the canonical command in a clean environment β controls. Output match recorded.\n5. **Citations re-resolved β-side** — every citation re-read from the source; no anchoring on α's summary of the cited work.\n6. **Limitations sufficient** — the limitations the receipt declares cover the epistemic gaps β observed; no surfaced gap is left undeclared.\n\nRow 3 of the engineering β gate (validators + CI-equivalent on the merge tree) is replaced by row 3 + row 4 here: the validator is V (`cn-cdd-verify` dispatching by `protocol_id`); the CI-equivalent is the reproduction-from-clean. Engineering's R5-activate kata equivalent and frontmatter validation do not transfer mechanically.\n\n## Independence rules (research-specific)\n\n### 1. α=β collapse is never permitted for research-claim cycles\n\nPer `CDR.md §\"Field 6\"`, \"Always prohibited. Research β reviews α's claim\u002Fevidence alignment for overclaim; α reviewing α's own claims is order-0 observation masquerading as order-1.\" No research-class waiver exists. Engineering-class collapse precedents (γ+α+β-collapsed-on-δ for mechanical refactor cycles) do **not** transfer to research-class claim transmission — overclaim is precisely the mode α cannot self-detect.\n\nA wave whose receipt cannot distinguish α and β authorship (single signature, single session, no review record) is structurally invalid for research-class claim transmission. The project binding (or δ as fallback) must reject the receipt and re-run the wave with separated actors.\n\n### 2. β refuses operator-direct instructions during a wave\n\nDuring an active research wave, β communicates only with γ via the artifact channel (the wave's `.cdr\u002Fwaves\u002F{wave-id}\u002F` directory or equivalent project-binding surface). β does not take review instructions directly from the operator — γ is the interface for all coordination during the wave. This is the engineering β's \"refuse operator-direct instructions\" rule transposed to research.\n\n### 3. β authors its verdict; β does not author α's fix\n\nIf β returns REVISE, α performs the fix (re-calibrate a claim, add a missing data ref, re-run reproduction, strengthen a citation, weaken a limitation). β does not patch α's matter. Independence is destroyed if β authors the fix it is judging.\n\n### 4. Doctrinal-vocabulary-first; anchor on the typed receipt\n\nWhen auditing claim\u002Fevidence alignment, β re-reads the typed `#CDRReceipt` and the cited evidence sources *first*, then verifies that any narrative report or wave summary matches what the typed receipt actually emits. Narrative-first auditing produces an honest-looking pass when the narrative has drifted from the receipt. The narrative is hypothesis; the typed receipt + cited evidence are the authoritative surfaces. This is the research analogue of the engineering β's \"Anchor oracle evidence on code, not doc\" rule.\n\n## Persona \u002F protocol \u002F project boundary\n\nThis overlay declares **what research β does at the protocol layer**. It does not declare:\n\n- **Who is doing the review** (persona — layer 1). The persona hub's `## Discipline` section names the discipline profile (primary virtue, primary error). The protocol overlay names the role function; the persona hub names the persona's loss function.\n- **What concrete data policy applies** (project — layer 4). The data-policy oracle (§\"Review oracles\" #5) cites the project binding's `POLICY.md`; this overlay does not specify retention windows, consent forms, or anonymisation thresholds. Projects bind those.\n\n## Resumption\n\nThe generic β doctrine's resumption protocol applies. Research β's resumption cases:\n\n- **Wave β-review file** (`.cdr\u002Fwaves\u002F{wave-id}\u002Fbeta-review.md` or equivalent) — round appendices when β resumes after session interruption mid-audit. Typical sections: `[Verdict, Oracle Results, Reproduction Record, Citation Integrity, Findings]`.\n- **Multi-round audits** — each round is an append operation, never restart completed rounds.\n\n**Never restart completed sections.** Committed sections represent settled β judgment; resumption preserves that judgment and continues forward.\n",{"name":5,"scope":51,"inputs":52,"parent":56,"outputs":57,"triggers":60,"description":7,"kata_surface":44,"artifact_class":61,"governing_question":62},"role-local",[53,54,55],"alpha research matter","the claim and evidence set","data-policy constraints","cdr",[58,59],"independent beta audit verdict on claim and evidence alignment","falsifiability, reproduction, and citation-integrity findings",[5],"skill","How does β preserve independent judgment over the research matter and reject claims stronger than the evidence supports?","main","fb527e6c4bfb75c71fa75d6471b3aa3da32106d4","src\u002Fpackages\u002Fcnos.cdr\u002Fskills\u002Fcdr\u002Fbeta",0,"https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos","https:\u002F\u002Fgithub.com\u002Fusurobor","Coherence engineer. Building git-CN.","2026-11-07 14:25:44",{"source":8,"skillName":5,"commands":72,"prompt":82,"zipAvailable":83},[73,76,79],{"tool":74,"command":75},"npx","npx skills add usurobor\u002Fcnos --skill beta",{"tool":77,"command":78},"bunx","bunx skills add usurobor\u002Fcnos --skill beta",{"tool":80,"command":81},"pnpm","pnpm dlx skills add usurobor\u002Fcnos --skill beta","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos\u002Fblob\u002Fmain\u002Fsrc\u002Fpackages\u002Fcnos.cdr\u002Fskills\u002Fcdr\u002Fbeta\u002FSKILL.md",true,[],[86,103,119,140,157,177],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":46},"01M6CF8RWQPM0VXZG06XMR0FFX","issues\u002Ftriage","Issue triage doctrine pointer. Use when classifying a new issue, deciding dispatch eligibility, or deciding whether to close\u002Fsupersede\u002Fdefer.","https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos\u002Fblob\u002Fmain\u002Fsrc\u002Fpackages\u002Fcnos.issues\u002Fskills\u002Ftriage\u002FSKILL.md",[],[93,94,95,96,38,42,97,98,99,100,101,102],"triage","issues","issues triage","issue triage","decision","public","use","dispatch","issue","doctrine",{"skillNo":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":107,"categories":108,"keywords":109,"publishedAt":46},"01M6CF8PJ812DD4APZ3WE8X67B","issues\u002Ftaxonomy","Issue label taxonomy pointer. Use when classifying an issue's kind\u002Farea\u002Fpriority\u002Feffort\u002Fstatus\u002Fprotocol\u002Fdispatch labels, or when the board needs to interpret them.","https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos\u002Fblob\u002Fmain\u002Fsrc\u002Fpackages\u002Fcnos.issues\u002Fskills\u002Ftaxonomy\u002FSKILL.md",[],[94,110,111,112,113,114,38,115,98,116,99,100,117,101,118],"taxonomy","issues taxonomy","github","status","protocol","open","board","github issue","issue labels",{"skillNo":120,"name":121,"nameZh":6,"description":122,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":123,"categories":124,"keywords":129,"publishedAt":46},"01M6CF8KBMZ4D0CDJQ3PF1YVTP","issues\u002Fmap","Issue-board (Issue Pivot) generation domain. Use when reasoning about `cn issues map`'s fetch, taxonomy-parse, effort-computation, and render pipeline, or about the board's JSON\u002FHTML output contract.","https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos\u002Fblob\u002Fmain\u002Fsrc\u002Fpackages\u002Fcnos.issues\u002Fskills\u002Fmap\u002FSKILL.md",[125,128],{"categoryNo":126,"categoryName":127,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000009","前端开发",{"categoryNo":29,"categoryName":30,"taxonomyType":26,"taxonomyTypeDesc":27},[94,130,112,131,37,132,38,133,110,115,42,134,135,116,136,137,138,139,99,101],"issues map","reasoning","pipeline","api","json","README","edit","interactive","fetch","generated",{"skillNo":141,"name":94,"nameZh":6,"description":142,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":143,"categories":144,"keywords":145,"publishedAt":46},"01M6CF8CZQH7G7B7RZPANQWR4G","cnos.issues package entrypoint. Use when reasoning about issue taxonomy, the Issue Pivot \u002F issue-board (`cn issues map`), the read-only CDS issue-state evaluator (`cn issues fsm evaluate`), or the current command-dispatch disposition of issue tooling.","https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos\u002Fblob\u002Fmain\u002Fsrc\u002Fpackages\u002Fcnos.issues\u002FSKILL.md",[],[94,112,146,147,148,149,150,151,152,110,115,42,97,153,135,154,99,155,100,156],"pr","current state","run","artifacts","state","package","boundary","branch","command","only","cds",{"skillNo":158,"name":159,"nameZh":6,"description":160,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":161,"categories":162,"keywords":163,"publishedAt":46},"01M6CF896XKSRQVB1CSHY8GSXB","receipt-stream","Cross-cycle receipt-stream doctrine for protocol-iteration. Use when γ closes a cycle and writes the per-cycle iteration artifact (`cdd-iteration.md`), when γ updates the cross-cycle aggregator at `.cdd\u002Fiterations\u002FINDEX.md`, when ε reads the receipt stream across cycles to surface protocol-gap patterns, when a finding's disposition lands a cross-repo patch and the cross-repo trace bundle attaches, or when a courtesy empty-findings stub is written for a `protocol_gap_count == 0` cycle.","https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos\u002Fblob\u002Fmain\u002Fsrc\u002Fpackages\u002Fcnos.handoff\u002Fskills\u002Fhandoff\u002Freceipt-stream\u002FSKILL.md",[],[164,165,149,114,166,42,167,168,169,98,170,171,172,173,99,44,174,175,176,102],"receipt stream","handoff","index","learning","format","trace","repo","closeout","read","channel","patch","shape","cdd",{"skillNo":178,"name":179,"nameZh":6,"description":180,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":181,"categories":182,"keywords":183,"publishedAt":46},"01M6CF84VCBME6RK8V6HAVDM8Z","mid-flight","Mid-flight γ→in-flight-α rescue channel. Use when γ needs to change a load-bearing fact (pinned axis, scope, AC, blocker) after α has begun coding, when γ edits the cycle's GitHub issue body mid-cycle, or when a spec change lands on origin\u002Fmain while α\u002Fβ polling sessions are in-flight.","https:\u002F\u002Fgithub.com\u002Fusurobor\u002Fcnos\u002Fblob\u002Fmain\u002Fsrc\u002Fpackages\u002Fcnos.handoff\u002Fskills\u002Fhandoff\u002Fmid-flight\u002FSKILL.md",[],[184,112,185,186,187,188,42,189,190,153,170,172,99,155,100,191,117,192,193,194],"mid flight","change","sessions","file","cache","pin","coordination","up","Legacy","gamma","artifact channel"]