Add executive summary claim trace assistant

executive-summary-claim-trace-assistant/README.md

@@ -0,0 +1,38 @@
+# Executive Summary Claim Trace Assistant
+
+This dependency-free module reviews AI-generated executive summaries before they
+are shared with collaborators, funders, or editors. It checks whether summary
+claims are traceable to manuscript evidence and whether implications or next
+steps are framed within the available evidence.
+
+## Checks
+
+- Each executive summary bullet includes at least one source evidence ID.
+- Referenced evidence IDs exist in the manuscript evidence packet.
+- Key findings are backed by results, tables, figures, or methods evidence.
+- Implications and next steps avoid unsupported universal language.
+- Clinical, policy, or funding claims require appropriate evidence types.
+- Numeric claims must match numbers present in the traced evidence.
+- Key findings must share enough terminology with the cited evidence to avoid
+  decorative or unrelated citations.
+- Unsupported or over-broad bullets receive reviewer rewrite suggestions.
+
+## Run
+
+```bash
+python3 executive-summary-claim-trace-assistant/summary_claim_trace_assistant.py \
+  --sample \
+  --json executive-summary-claim-trace-assistant/demo/report.json \
+  --markdown executive-summary-claim-trace-assistant/demo/summary.md \
+  --svg executive-summary-claim-trace-assistant/demo/graph.svg
+```
+
+Generated demo artifacts in `demo/` show the JSON decision payload, a reviewer
+Markdown table, an SVG status chart, and a short MP4 walkthrough for bounty
+reviewers.
+
+## Test
+
+```bash
+python3 -m unittest executive-summary-claim-trace-assistant/test_summary_claim_trace_assistant.py
+```

executive-summary-claim-trace-assistant/demo/report.json

@@ -0,0 +1,134 @@
+{
+  "assistant": "executive-summary-claim-trace-assistant",
+  "bullet_reviews": [
+    {
+      "bullet_id": "SUM-001",
+      "decision": "publish",
+      "evidence_trace": [
+        {
+          "evidence_id": "EV-RESULT-001",
+          "section": "Results",
+          "type": "result"
+        }
+      ],
+      "findings": [],
+      "mode": "key_finding",
+      "rewrite": ""
+    },
+    {
+      "bullet_id": "SUM-002",
+      "decision": "publish",
+      "evidence_trace": [
+        {
+          "evidence_id": "EV-RESULT-001",
+          "section": "Results",
+          "type": "result"
+        },
+        {
+          "evidence_id": "EV-FIG-002",
+          "section": "Figure 2",
+          "type": "figure"
+        },
+        {
+          "evidence_id": "EV-METHOD-003",
+          "section": "Methods",
+          "type": "method"
+        }
+      ],
+      "findings": [],
+      "mode": "implication",
+      "rewrite": ""
+    },
+    {
+      "bullet_id": "SUM-003",
+      "decision": "revise",
+      "evidence_trace": [
+        {
+          "evidence_id": "EV-RESULT-001",
+          "section": "Results",
+          "type": "result"
+        }
+      ],
+      "findings": [
+        {
+          "code": "OVERBROAD_LANGUAGE",
+          "message": "Summary uses universal or definitive language beyond the evidence.",
+          "severity": "revise"
+        },
+        {
+          "code": "CLINICAL_CLAIM_NEEDS_SAFETY_CONTEXT",
+          "message": "Clinical claims need methods, ethics, trial, or limitation evidence.",
+          "severity": "revise"
+        }
+      ],
+      "mode": "next_step",
+      "rewrite": "The tool could be evaluated across similar clinical publishing workflows."
+    },
+    {
+      "bullet_id": "SUM-004",
+      "decision": "block",
+      "evidence_trace": [],
+      "findings": [
+        {
+          "code": "UNKNOWN_EVIDENCE_ID",
+          "message": "Summary references evidence IDs that are absent from the packet: EV-MISSING-999",
+          "severity": "block"
+        }
+      ],
+      "mode": "key_finding",
+      "rewrite": "Hold this bullet until the source evidence IDs are added or corrected."
+    },
+    {
+      "bullet_id": "SUM-005",
+      "decision": "block",
+      "evidence_trace": [],
+      "findings": [
+        {
+          "code": "MISSING_EVIDENCE_TRACE",
+          "message": "Summary bullet has no source evidence IDs.",
+          "severity": "block"
+        },
+        {
+          "code": "OVERBROAD_LANGUAGE",
+          "message": "Summary uses universal or definitive language beyond the evidence.",
+          "severity": "revise"
+        }
+      ],
+      "mode": "implication",
+      "rewrite": "Hold this bullet until the source evidence IDs are added or corrected."
+    },
+    {
+      "bullet_id": "SUM-006",
+      "decision": "revise",
+      "evidence_trace": [
+        {
+          "evidence_id": "EV-RESULT-001",
+          "section": "Results",
+          "type": "result"
+        }
+      ],
+      "findings": [
+        {
+          "code": "UNSUPPORTED_NUMERIC_DETAIL",
+          "message": "Summary includes numeric details absent from traced evidence: 50",
+          "severity": "revise"
+        },
+        {
+          "code": "OVERBROAD_LANGUAGE",
+          "message": "Summary uses universal or definitive language beyond the evidence.",
+          "severity": "revise"
+        }
+      ],
+      "mode": "key_finding",
+      "rewrite": "Replace unsupported numbers with the exact values in the traced evidence or remove the numeric detail."
+    }
+  ],
+  "manuscript_id": "MS-EXEC-SUMMARY-2026",
+  "summary": {
+    "block": 2,
+    "publish": 2,
+    "revise": 2,
+    "total": 6,
+    "trace_count": 6
+  }
+}

executive-summary-claim-trace-assistant/demo/summary.md

@@ -0,0 +1,17 @@
+# Executive Summary Claim Trace Report
+
+- Manuscript: MS-EXEC-SUMMARY-2026
+- Total bullets: 6
+- Publish: 2
+- Revise: 2
+- Block: 2
+- Evidence traces: 6
+
+| Bullet | Decision | Findings | Rewrite |
+| --- | --- | --- | --- |
+| SUM-001 | publish | none |  |
+| SUM-002 | publish | none |  |
+| SUM-003 | revise | OVERBROAD_LANGUAGE, CLINICAL_CLAIM_NEEDS_SAFETY_CONTEXT | The tool could be evaluated across similar clinical publishing workflows. |
+| SUM-004 | block | UNKNOWN_EVIDENCE_ID | Hold this bullet until the source evidence IDs are added or corrected. |
+| SUM-005 | block | MISSING_EVIDENCE_TRACE, OVERBROAD_LANGUAGE | Hold this bullet until the source evidence IDs are added or corrected. |
+| SUM-006 | revise | UNSUPPORTED_NUMERIC_DETAIL, OVERBROAD_LANGUAGE | Replace unsupported numbers with the exact values in the traced evidence or remove the numeric detail. |

executive-summary-claim-trace-assistant/sample_summary_packet.json

@@ -0,0 +1,84 @@
+{
+  "manuscript_id": "MS-EXEC-SUMMARY-2026",
+  "evidence": [
+    {
+      "id": "EV-RESULT-001",
+      "type": "result",
+      "section": "Results",
+      "text": "The screened model reduced annotation time from 42 minutes to 31 minutes across 84 benchmark manuscripts.",
+      "metrics": {
+        "baseline_minutes": 42,
+        "intervention_minutes": 31,
+        "n": 84
+      }
+    },
+    {
+      "id": "EV-FIG-002",
+      "type": "figure",
+      "section": "Figure 2",
+      "text": "Figure 2 shows lower reviewer disagreement for biomedical manuscripts after the checklist prompt was enabled."
+    },
+    {
+      "id": "EV-METHOD-003",
+      "type": "method",
+      "section": "Methods",
+      "text": "Evaluation used a retrospective corpus and did not include prospective clinical deployment."
+    },
+    {
+      "id": "EV-LIMIT-004",
+      "type": "limitation",
+      "section": "Limitations",
+      "text": "The study was conducted in English-language manuscripts only."
+    }
+  ],
+  "summary_bullets": [
+    {
+      "id": "SUM-001",
+      "mode": "key_finding",
+      "text": "The assistant reduced annotation time by roughly one quarter in the benchmark set.",
+      "evidence_ids": [
+        "EV-RESULT-001"
+      ]
+    },
+    {
+      "id": "SUM-002",
+      "mode": "implication",
+      "text": "These results suggest the workflow may reduce reviewer triage friction for similar manuscript review settings.",
+      "evidence_ids": [
+        "EV-RESULT-001",
+        "EV-FIG-002",
+        "EV-METHOD-003"
+      ]
+    },
+    {
+      "id": "SUM-003",
+      "mode": "next_step",
+      "text": "The tool should now be deployed across all clinical publishing workflows.",
+      "evidence_ids": [
+        "EV-RESULT-001"
+      ]
+    },
+    {
+      "id": "SUM-004",
+      "mode": "key_finding",
+      "text": "The assistant works equally well for multilingual manuscripts.",
+      "evidence_ids": [
+        "EV-MISSING-999"
+      ]
+    },
+    {
+      "id": "SUM-005",
+      "mode": "implication",
+      "text": "This will eliminate peer-review bottlenecks.",
+      "evidence_ids": []
+    },
+    {
+      "id": "SUM-006",
+      "mode": "key_finding",
+      "text": "The assistant reduced workload by 50% across all manuscripts.",
+      "evidence_ids": [
+        "EV-RESULT-001"
+      ]
+    }
+  ]
+}