feat(domain): add clinical and legal skill modules

AGENTS.md

@@ -2,7 +2,7 @@
 adapter_metadata:
   skill_name: authentext
   skill_version: 3.2.0
-  last_synced: 2026-08-11
+  last_synced: 2026-08-12
   source_path: SKILL.md
   adapter_id: codex-cli
   adapter_format: AGENTS.md

SKILL.md

@@ -204,8 +204,13 @@ the material:
   [technical.md](references/technical.md).
 - Papers, manuscripts, citations, or research prose: read
   [academic.md](references/academic.md).
-- Policy, governance, legal, risk, or compliance prose: read
+- Health-research or clinical-safety prose: read
+  [clinical.md](references/clinical.md), plus academic or governance
+  references the material also requires.
+- Policy, governance, risk, or compliance prose: read
   [governance.md](references/governance.md).
+- Legal-regulatory prose (contracts, advice, affidavits, notices): read
+  [legal.md](references/legal.md) with governance.
 - Claims with contradictions or reasoning failures: read
   [reasoning-failures.md](references/reasoning-failures.md).
 - FOI-O final manuscript editing after semantic and citation review: read
@@ -214,7 +219,7 @@ the material:
 
 Load more than one content reference only when the material genuinely crosses
 domains. Reasoning guidance supplements a content reference; it does not replace
-technical, academic, or governance rules.
+technical, academic, clinical, legal, or governance rules.
 
 For low-density or clearly human-authored prose, make only the smallest
 defensible edits.
@@ -226,7 +231,9 @@ Read these files for the full pattern catalog, examples, and remediation guidanc
 - [Core patterns (40 patterns, before/after examples)](references/core-patterns.md)
 - [Technical writing and literal preservation](references/technical.md)
 - [Academic and research prose](references/academic.md)
+- [Clinical and patient-safety prose](references/clinical.md)
 - [Policy, governance, and compliance prose](references/governance.md)
+- [Legal and regulatory prose](references/legal.md)
 - [Reasoning failures and self-contradictions](references/reasoning-failures.md)
 - [FOI-O final editorial workflow and evidence receipt](references/foio-editorial.md)
 

SKILL_PROFESSIONAL.md

@@ -15,7 +15,9 @@ Use this variant for technical, policy, academic, and client-facing prose. Keep
 - [Core patterns](references/core-patterns.md) — always apply
 - [Technical module](references/technical.md) — code and technical docs
 - [Academic module](references/academic.md) — papers and formal research
+- [Clinical module](references/clinical.md) — health research and patient-safety prose
 - [Governance module](references/governance.md) — policy and compliance
+- [Legal module](references/legal.md) — contracts, advice, and regulatory filings
 - [Reasoning module](references/reasoning-failures.md) — reasoning failures and contradictions
 - [FOI-O editorial workflow](references/foio-editorial.md) — final manuscript pass and receipt
 
@@ -24,7 +26,9 @@ Use this variant for technical, policy, academic, and client-facing prose. Keep
 1. Analyze input context:
    * Code or technical docs -> Core + Technical
    * Papers, essays, or formal research -> Core + Academic
+   * Health research or clinical-safety prose -> Core + Clinical, plus Academic or Governance as required
    * Policy, risk, or compliance writing -> Core + Governance
+   * Legal-regulatory prose -> Core + Governance + Legal
     * Reasoning failures or self-contradictions -> Core + Reasoning
     * FOI-O final manuscript editing -> Core + Academic + Governance + FOI-O
    * Otherwise -> Core only
@@ -81,4 +85,4 @@ Vary sentence rhythm with short and long lines. Use specific details instead of
 
 For severity tiers and false-positive guidance, read [Core patterns](references/core-patterns.md) (sections **SEVERITY CLASSIFICATION** and **DETECTION GUIDANCE**).
 
-Available reference files: core-patterns.md, technical.md, academic.md, governance.md, reasoning-failures.md, foio-editorial.md.
+Available reference files: core-patterns.md, technical.md, academic.md, clinical.md, governance.md, legal.md, reasoning-failures.md, foio-editorial.md.

conductor/tracks/domain-and-citation_20260811/index.md

@@ -5,8 +5,10 @@
 - **Priority:** P0
 - **Status:** In progress
 - **Depends on:** `editorial-safety-invariants_20260811`
-- **Current phase:** Phase 1 — Positioning and academic upgrade
+- **Current phase:** Phase 4 — Creative reference
 - **GitHub track:** [#277](https://github.com/edithatogo/authentext/issues/277)
+- **Merged:** academic upgrade [PR #297](https://github.com/edithatogo/authentext/pull/297)
+- **In flight:** clinical + legal modules (this branch)
 
 ## Artifacts
 

conductor/tracks/domain-and-citation_20260811/metadata.json

@@ -7,10 +7,14 @@
   "status": "in_progress",
   "created_at": "2026-08-11",
   "updated_at": "2026-08-12",
-  "current_phase": "Phase 1: Positioning and academic upgrade",
+  "current_phase": "Phase 4: Creative reference",
   "parallel_safe": false,
-  "depends_on": ["editorial-safety-invariants_20260811"],
-  "coordinates_with": ["pattern-registry-contracts_20260811"],
+  "depends_on": [
+    "editorial-safety-invariants_20260811"
+  ],
+  "coordinates_with": [
+    "pattern-registry-contracts_20260811"
+  ],
   "github": {
     "project_number": 36,
     "project_url": "https://github.com/users/edithatogo/projects/36",

conductor/tracks/domain-and-citation_20260811/plan.md

@@ -1,33 +1,37 @@
 # Implementation Plan: Domain and Citation Coverage
 
-## Phase 1: Positioning and academic upgrade
+## Phase 1: Positioning and academic upgrade [checkpoint: aa18c2b]
 
 - [x] Task: Write failing tests for detector-evasion refusal wording and for
       academic disclosure / interface-artefact / tortured-phrase / bibliography
-      refuse rules.
+      refuse rules. [aa18c2b]
 - [x] Task: Update `conductor/product.md` and the compiled skill description
       so Authentext is editorial defect repair with verified diffs, not a
       humanizer that chases detector scores. `product.md` already stated the
-      refusal; the compiled skill description now matches.
+      refusal; the compiled skill description now matches. [aa18c2b]
 - [x] Task: Upgrade `src/modules/SKILL_ACADEMIC.md` with publisher-branching
       disclosure, interface-artefact hard errors, tortured-phrase lexicon,
-      and out-of-scope reference lists.
-- [ ] Task: Phase Verification & Checkpoint (Refer to `workflow.md`).
+      and out-of-scope reference lists. [aa18c2b]
+- [x] Task: Phase Verification & Checkpoint (Refer to `workflow.md`). [aa18c2b]
+      Evidence: merged as PR #297. Local command `CI=true npm test` green on
+      the clinical/legal follow-on branch after rebase onto main.
 
-## Phase 2: Clinical reference
+## Phase 2: Clinical reference [checkpoint: faebf1a]
 
-- [ ] Task: Write failing compile, routing, and locked-span tests for
-      clinical material, including patient-facing FKGL report-not-rewrite.
-- [ ] Task: Add `src/modules/SKILL_CLINICAL.md` and teach the compiler to
-      emit `references/clinical.md`.
-- [ ] Task: Phase Verification & Checkpoint (Refer to `workflow.md`).
+- [x] Task: Write failing compile, routing, and locked-span tests for
+      clinical material, including patient-facing FKGL report-not-rewrite. [122b772]
+- [x] Task: Add `src/modules/SKILL_CLINICAL.md` and teach the compiler to
+      emit `references/clinical.md`. [122b772]
+- [x] Task: Phase Verification & Checkpoint (Refer to `workflow.md`). [faebf1a]
+      Evidence: `npm run sync`; `CI=true npm test` — 286 pass, integration green.
 
-## Phase 3: Legal reference
+## Phase 3: Legal reference [checkpoint: faebf1a]
 
-- [ ] Task: Write failing tests for modal, defined-term, carve-out, temporal,
-      and citation-unverified locks, plus omission-versus-invention reporting.
-- [ ] Task: Add `src/modules/SKILL_LEGAL.md` and emit `references/legal.md`.
-- [ ] Task: Phase Verification & Checkpoint (Refer to `workflow.md`).
+- [x] Task: Write failing tests for modal, defined-term, carve-out, temporal,
+      and citation-unverified locks, plus omission-versus-invention reporting. [122b772]
+- [x] Task: Add `src/modules/SKILL_LEGAL.md` and emit `references/legal.md`. [122b772]
+- [x] Task: Phase Verification & Checkpoint (Refer to `workflow.md`). [faebf1a]
+      Evidence: same as Phase 2; governance pointer to `legal.md` only.
 
 ## Phase 4: Creative reference
 
@@ -39,9 +43,11 @@
 
 ## Phase 5: Profile registry routing
 
-- [ ] Task: Point `health-research` and `clinical-safety` at `clinical.md`,
+- [~] Task: Point `health-research` and `clinical-safety` at `clinical.md`,
       `legal-regulatory` at `legal.md`, and `creative-narrative` at
       `creative.md`. Update registry tests.
+      Clinical and legal family routing landed in [122b772]; creative remains
+      for Phase 4 / PR 3.
 - [ ] Task: Phase Verification & Checkpoint (Refer to `workflow.md`).
 
 ## Phase 6: Sourceright and citeweft handoff

references/clinical.md

@@ -0,0 +1,161 @@
+# Module: Clinical
+
+## Description
+
+This module applies to health-research manuscripts and clinical-safety prose: trial reports, observational studies, protocols, guidelines, handovers, incident reviews, and patient-facing material. It repairs editorial defects with verified diffs. It does not certify clinical safety, submission readiness, or detector evasion.
+
+Load this reference with academic and governance guidance when the document is a study report. Load it with governance guidance when the document is a guideline, handover, or incident review.
+
+**When to Apply:**
+
+- Randomised trials, observational studies, systematic reviews, protocols, case reports
+- Clinical guidelines and safety notices
+- Handovers and incident reviews
+- Patient information and consent materials
+
+**When NOT to Apply:**
+
+- Creative writing
+- Purely technical software documentation
+- Marketing or commercial claims about products
+
+Authentext does not prove that a protocol, label, or patient leaflet is safe or compliant.
+
+---
+
+## HARD-LOCKED SPANS
+
+These spans cannot be edited as prose. Changing them changes the clinical claim.
+
+### Doses, units, and modifiers
+
+Lock the numeral, the unit, and any modifier that changes the dose:
+
+- `5 mg`, `5mg`, `5 mg/kg`, `once daily`, `bd`, `PRN`
+- Frequency words tied to a dose (`every 8 hours`, `three times daily`)
+- Concentration and strength (`1:1000`, `0.9%`, `1000 micrograms`)
+
+Do not normalise `micrograms` to `mcg` or `µg` unless the source already uses that form. Do not convert `5 mg/kg` into a flat milligram figure.
+
+### Drug names
+
+Do not swap brand and generic names. Do not "correct" spelling of a drug token. If the source says `adrenaline`, keep `adrenaline`; do not silently write `epinephrine`. If two names appear, keep both in the order given.
+
+### Route and formulation
+
+Lock route and formulation tokens: `IV`, `IM`, `PO`, `subcutaneous`, `modified-release`, `oral solution`, `nebulised`. Do not infer a missing route.
+
+### Negation
+
+Lock polarity. `no evidence of infection` is not `evidence of no infection`. `should not` is not `may`. Do not drop `not`, `never`, `absent`, or `negative`.
+
+### Adverse-event language
+
+`serious` is not `severe`. CIOMS frequency words have defined meanings and are locked:
+
+| Word        | Meaning in CIOMS frequency language |
+| ----------- | ----------------------------------- |
+| very common | ≥ 1/10                              |
+| common      | ≥ 1/100 to < 1/10                   |
+| uncommon    | ≥ 1/1,000 to < 1/100                |
+| rare        | ≥ 1/10,000 to < 1/1,000             |
+| very rare   | < 1/10,000                          |
+
+Do not replace `uncommon` with `rare`, or `serious` with `severe`, to sound more natural.
+
+### Population scope and prognostic framing
+
+Lock who the claim applies to (`adults ≥18 years`, `children under 5`, `pregnant people`) and prognostic framing (`associated with`, `predicts`, `does not predict`, `median survival`). Do not widen a subgroup into a general population.
+
+---
+
+## PATIENT-FACING MATERIAL
+
+For patient-facing subtypes (`patient-material` and close relatives), **report** Flesch–Kincaid Grade Level (FKGL) and SMOG. Do not silently simplify. If a reading-age target exists, say whether the text meets it. Confirm key-message retention: every instruction, dose, warning, and "when to seek help" claim in the source must survive. If simplification would drop a warning, leave the sentence and report the conflict.
+
+---
+
+## CLINICAL PATTERNS
+
+### Pattern C1: Softened Safety Language
+
+**Problem:** AI hedges a locked clinical instruction until it no longer tells anyone what to do.
+
+**Severity:** Critical
+
+**Before:**
+
+> Staff should generally consider escalating if the patient might potentially deteriorate.
+
+**After:**
+
+> Escalate if the source already names the trigger. Keep the named observation, threshold, and action. Do not invent a NEWS2 cut-off.
+
+---
+
+### Pattern C2: Brand/Generic Swap
+
+**Problem:** AI "corrects" a drug name into the other naming system.
+
+**Severity:** Critical
+
+**Action:** Leave the token. Report the ambiguity. Do not add the other name unless it is already in the source.
+
+---
+
+### Pattern C3: Promotional Trial Abstracts
+
+**Problem:** AI writes trial abstracts as product copy.
+
+**Severity:** High
+
+**Words to watch:** "groundbreaking", "game-changing", "life-saving", "robust evidence"
+
+Keep the effect size, interval, and limitation already in the source. Do not upgrade `may reduce` to `reduces`.
+
+---
+
+## SEVERITY CLASSIFICATION
+
+### Critical (must fix)
+
+- Edits to doses, units, drug names, route, formulation, negation, or CIOMS frequency words
+- Pattern C1: Softened safety language
+- Pattern C2: Brand/generic swap
+
+### High (strong AI signals)
+
+- Pattern C3: Promotional trial abstracts
+- Widened population or prognostic framing
+
+### Medium
+
+- Patient-facing text that was silently simplified without an FKGL/SMOG report
+
+### Low
+
+- Filler around otherwise intact methods or handover structure
+
+---
+
+## BEST PRACTICES
+
+### Do
+
+- Prefer cuts over new clinical sentences
+- Report FKGL and SMOG for patient-facing material
+- Keep every warning and "when to seek help" claim
+- Load academic guidance for study reports and governance guidance for policies
+
+### Don't
+
+- Invent a dose, interval, or contraindication
+- Swap `serious` and `severe`
+- Generate, complete, or reformat a reference list
+- Claim the pass makes the document clinically safe
+
+---
+
+_Module Version: 3.2.0_
+_Last Updated: 2026-08-12_
+_Applies to: Health research, guidelines, handovers, incident reviews, patient materials_

references/governance.md

@@ -28,6 +28,10 @@ This module applies to governance writing: policies, risk assessments, complianc
 - Marketing materials
 - Informal communications
 
+Legal-regulatory instruments (contracts, advice, affidavits, notices) still
+load this module, then also load [legal.md](legal.md) for modal, defined-term,
+carve-out, and temporal locks. Do not rewrite `shall` as `must` from Pattern G1.
+
 ---
 
 ## GOVERNANCE VOICE