zrwangems/dsh-zrwang-plugin / ars/skills/sr-screener
sr-screener
Flagged系统综述文献筛选:把方案或 PROSPERO 协议转成资格规则,双盲筛选标题摘要与全文,第三人裁决冲突,输出 PRISMA 2020 计数、EndNote/Zotero RIS 分组与可交付给论文技能的 literature_corpus。Triggers: systematic review screening, screen papers, PRISMA counts, pilot screening, adjudicate conflicts | 文献筛选, 系统综述筛选, PRISMA 计数, 筛选冲突裁决。
_shared/contracts/passport/experiment_alignment_result.schema.json
View on GitHub{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"$id": "https://github.com/Imbad0202/academic-research-skills/shared/contracts/passport/experiment_alignment_result.schema.json",
"title": "Material Passport Experiment Alignment Result Entry",
"description": "One entry in Material Passport Schema 9 experiment_alignment_results[] aggregate (#260) — the FOURTH ref_slug-less claim-finding aggregate, alongside uncited_assertions[] / claim_drifts[] / constraint_violations[]. An experiment-backed claim has no <!--ref:slug--> citation marker, so it cannot satisfy claim_audit_result.required (which mandates ref_slug); it rides its own aggregate by the same precedent the three siblings established. PRODUCER = the integrity verification agent, computed AT the integrity gate (Stage 2.5 sampling / Stage 4.5 full) — NOT the claim_ref_alignment_audit_agent. This mirrors #261's Phase C3 (figure-fidelity verdict computed by the integrity agent at the gate, not pre-computed upstream): producing the verdict at the same stage that blocks on it avoids the stage-ordering race where a verdict produced at the Stage 4->5 boundary would land AFTER the Stage 4.5 gate already ran. A claim referencing an experiment_id with NO matching experiment_provenance[] entry is a STRUCTURAL failure (EP-INV-2 / EA-INV-2), never a judge verdict — PROVENANCE_MISSING is deliberately NOT in the verdict enum. Cross-array invariants EA-INV-1..EA-INV-2 are lint-enforced in scripts/check_claim_audit_consistency.py. See docs/design/2026-06-08-260-experiment-provenance-intake-spec.md D4.",
"type": "object",
"additionalProperties": false,
"required": [
"finding_id",
"scoped_manifest_id",
"claim_id",
"claim_text",
"experiment_id",
"result_pointer",
"manuscript_locator",
"alignment_verdict",
"rationale",
"judge_model",
"judge_run_at",
"rule_version"
],
"properties": {
"finding_id": {
"type": "string",
"pattern": "^EA-[0-9]{3,}$",
"description": "Per-passport unique finding identifier for experiment_alignment_results[]. EA-INV-1 enforces uniqueness. EA- prefix mirrors the UA-/DR-/CV- sibling-aggregate convention."
},
"scoped_manifest_id": {
"type": "string",
"pattern": "^M-[0-9]{4}-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}Z-[0-9a-f]{4}$",
"description": "Points to the owning claim_intent_manifest.manifest_id. The pair (scoped_manifest_id, claim_id) is the joinable key — bare claim_id collides across manifests in the same passport. EA-INV-2 enforces cross-array resolution to a real manifest claim."
},
"claim_id": {
"type": "string",
"pattern": "^C-[0-9]{3,}$",
"description": "Per-manifest claim identifier scoped by scoped_manifest_id."
},
"claim_text": { "type": "string", "minLength": 1, "maxLength": 2000 },
"experiment_id": {
"type": "string",
"minLength": 1,
"description": "The experiment_provenance[].experiment_id this claim is judged against. EA-INV-2 enforces resolution to a real experiment_provenance[] entry; a dangling id is a structural FAIL, never modelled as a verdict."
},
"result_pointer": {
"type": "string",
"minLength": 1,
"description": "Points INTO the provenance entry's reported results (e.g. a planned_vs_executed result_file + metric, or a JSON-Pointer-style locator) — experiment_id alone is too coarse for a claim like 'F1 improved 4.2%'."
},
"manuscript_locator": {
"type": "string",
"minLength": 1,
"description": "Section path to the claim sentence (e.g. '4. Results > 4.2 Ablations') so a failing alignment can be fixed in the manuscript."
},
"alignment_verdict": {
"enum": ["ALIGNED", "OVERSTATED", "NOT_SUPPORTED_BY_PROVENANCE", "PROVENANCE_INSUFFICIENT"],
"description": "MECE, experiment-specific verdict (deliberately NOT reusing claim_audit_result.judgment names to avoid two UNSUPPORTED semantics). ALIGNED = supported by declared provenance. OVERSTATED = provenance supports a weaker version (claim says more than results show). NOT_SUPPORTED_BY_PROVENANCE = experiment ran but results do not support the claim (no relevant result / contradicts / irrelevant / contradicts a declared negative_results[] entry / every planned_vs_executed[] entry for the referenced experiment is executed:false). PROVENANCE_INSUFFICIENT = the entry exists but lacks detail to judge. PROVENANCE_MISSING is NOT here — a dangling experiment_id is an EP-INV-2/EA-INV-2 structural FAIL."
},
"rationale": { "type": "string", "minLength": 1, "maxLength": 2000 },
"judge_model": { "type": "string", "minLength": 1 },
"judge_run_at": { "type": "string", "format": "date-time" },
"rule_version": {
"const": "EA-v1",
"description": "Frozen literal for the #260 release. Future rule revisions bump the constant and require re-lint."
}
}
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |