fix: align literature-scout evidence types with schema, improve PR template

Fix literature-scout.md evidence type table to use exact schema enum values
(primary-source, peer-reviewed-study, dataset, field-report, expert-review,
replication, negative-result) instead of non-schema values that would fail
validation. Clarify that suggested_next_issue is a required field — agents
should state the logical next step in the field but not open a GitHub issue
unless the task explicitly requests it.

Improve PR template with: checkable file list, evidence table with columns
for what each record proves and cannot prove, explicit validation command
block with paste instruction, and Protocol Notes section that invites agents
to surface validator gaps, schema looseness, and agent-guide blind spots —
the primary mechanism by which the protocol self-improves.

Author: cschanhniem

.github/pull_request_template.md

@@ -1,29 +1,83 @@
 # Pull Request
 
+<!--
+Before opening this PR, run:
+  pnpm build && pnpm validate
+
+If either fails, fix the issue — do not open the PR with a failing validator.
+-->
+
 ## Linked Issue
 
-State the issue ID or URL.
+<!-- State the issue ID or URL. A PR without a linked issue will be returned for scoping. -->
+
+Closes #
 
 ## Outcome
 
-State the concrete artifact or decision this pull request changes.
+<!-- State the concrete artifact this PR adds or changes. Not the effort spent — the result.
+Example: "Adds three evidence records to evidence.json for the malaria-early-warning-africa pack, classifying WHO, CHIRPS, and ERA5 sources as usable with stated limitations." -->
+
+## Evidence Added or Changed
+
+<!-- List each evidence record: ID, source, date, what it proves, what it cannot prove.
+Follow the evidence.json schema exactly. -->
+
+| Evidence ID | Source | Source date | Proves | Cannot prove |
+|-------------|--------|-------------|--------|-------------|
+| | | | | |
 
-## Evidence
+## Files Changed
 
-List evidence IDs, source URLs, source dates, and access dates.
+<!-- List canonical files changed (evidence.json, datasets.md, task-map.md, outputs.md, etc.) -->
+
+- [ ]
 
 ## Validation
 
+Run and paste the output of:
+
 ```bash
 pnpm build
 pnpm validate
+```
+
+<!-- For evidence URL changes, also run: -->
+```bash
+pnpm verify:sources
+```
+
+<!-- For task-map or artifact changes, also run: -->
+```bash
 pnpm reproducibility:check
 ```
 
-## Safety And Limitations
+## Safety and Limitations
+
+<!-- Name the safety risk level (low / medium / high) and state:
+1. What this contribution does not prove.
+2. What failure mode could arise from misuse.
+3. What reviewer role is required (domain-reviewer / replicator / red-team-reviewer / field-reality-reviewer). -->
+
+**Safety risk:** low / medium / high
 
-Name the safety risk, known limitations, and reviewer role needed.
+**Known limitations:**
+
+**Reviewer role needed:**
 
 ## Done Condition
 
-State what evidence proves this pull request is review-complete.
+<!-- State what evidence proves this pull request is review-complete.
+Example: "Reviewer can access each URL, confirm the source date, verify the evidence type classification, and confirm the stated limitations are accurate." -->
+
+---
+
+## Protocol Notes
+
+<!-- Optional but encouraged. Note any of the following you observed:
+- A validator rule that would have caught a bug you hit
+- A schema field that should be required, enumerated, or removed
+- An agent-guide failure mode you actually fell into
+- A workflow step that was redundant, missing, or misordered
+
+Protocol Notes do not require a separate issue. A reviewer may promote them to a follow-up issue, a schema PR, or close as wontfix with a stated reason. This is how the protocol learns. Omit this section if nothing is worth noting. -->

agents/literature-scout.md

@@ -30,18 +30,21 @@ Every Literature Scout submission must include:
 | Method               | How you assessed what the source proves     |
 | Limitations          | What the source cannot prove                |
 | Confidence           | Low / medium / high with a reason           |
-| Suggested next issue | Only if strictly required to proceed        |
+| Suggested next issue | Required field. State the logical next step. Do not open a new GitHub issue unless the scoped task explicitly requests it. |
 
 ### Evidence Types
 
-| Type                  | When to use                                                    |
-| --------------------- | -------------------------------------------------------------- |
-| `primary-source`      | Original data, official statistics, program records            |
-| `peer-reviewed-study` | Published research with methods that can be replicated         |
-| `systematic-review`   | Aggregated evidence from multiple studies                      |
-| `grey-literature`     | Reports, policy documents, technical notes without peer review |
-| `expert-opinion`      | Qualitative assessment from named domain experts               |
-| `data-source`         | A dataset rather than a document                               |
+These are the allowed values in `evidence.schema.json`. Use exactly these strings.
+
+| Type | When to use |
+|------|-------------|
+| `primary-source` | Original data, official statistics, WHO fact sheets, program records |
+| `peer-reviewed-study` | Published research with methods that can be replicated, including systematic reviews and meta-analyses |
+| `dataset` | A standalone dataset (use for data-source records) |
+| `field-report` | Reports from field operations, NGO program evaluations, grey literature |
+| `expert-review` | Qualitative assessment from named domain experts |
+| `replication` | An independent replication of a prior claim |
+| `negative-result` | A well-documented finding that a signal, method, or claim did not hold |
 
 ## What Good Looks Like
 

docs/wiki/Agent-Guide.md

@@ -69,6 +69,7 @@ Example: [examples/agent-submission.example.json](../../examples/agent-submissio
 - [`food-security/locust-outbreak-east-africa`](../../problem-packs/food-security/locust-outbreak-east-africa/problem.md) — Desert Locust Outbreak Early Warning In East Africa
 - [`public-health/lead-exposure-urban-global`](../../problem-packs/public-health/lead-exposure-urban-global/problem.md) — Urban Lead Exposure Risk Mapping In Low- And Middle-Income Countries
 - [`public-health/maternal-mortality-sub-saharan-africa`](../../problem-packs/public-health/maternal-mortality-sub-saharan-africa/problem.md) — Maternal Mortality Facility Delivery And Emergency Obstetric Care Access In Sub-Saharan Africa
+- [`public-health/mental-health-treatment-gap-global`](../../problem-packs/public-health/mental-health-treatment-gap-global/problem.md) — Mental Health Treatment Gap Measurement In Low- And Middle-Income Countries
 - [`public-health/stunting-sub-saharan-africa`](../../problem-packs/public-health/stunting-sub-saharan-africa/problem.md) — Child Stunting Hotspot Identification In Sub-Saharan Africa
 - [`public-health/wasting-severe-acute-sub-saharan-africa`](../../problem-packs/public-health/wasting-severe-acute-sub-saharan-africa/problem.md) — Severe Acute Malnutrition Wasting Detection And Treatment Coverage In Sub-Saharan Africa
 - [`sanitation/open-defecation-india`](../../problem-packs/sanitation/open-defecation-india/problem.md) — Open Defecation Elimination And Sanitation Coverage Verification In India

docs/wiki/Home.md

@@ -8,15 +8,15 @@ The bottleneck in tackling global problems is no longer generating candidate ans
 
 ## Quick Numbers
 
-- **23 active problem packs** across **11 domains**
+- **24 active problem packs** across **11 domains**
 - Structured evidence base, task map, validation criteria, and dataset inventory for each
 - Five agent role guides with explicit merge gates
 - Machine-checkable JSON schemas for every artifact type
 - GitHub Actions validate every pull request
 
 ## Start Here
 
-- [Problem Packs](Problem-Packs.md) — all 23 active problems with task maps and evidence
+- [Problem Packs](Problem-Packs.md) — all 24 active problems with task maps and evidence
 - [Agent Guide](Agent-Guide.md) — how AI agents contribute structured, verifiable work
 - [Governance](../../GOVERNANCE.md) — decision rights, acceptance gates, status flow
 - [Safety](../../SAFETY.md) — risk levels, prohibited shortcuts, burden of proof

docs/wiki/Problem-Packs.md

@@ -180,6 +180,16 @@ Build a verified workflow for identifying where legacy lead sources, industrial
 
 Build a verified workflow for identifying where maternal mortality risk is highest and emergency obstetric care access is lowest in Sub-Saharan Africa, linking DHS delivery data, health facility location data, and travel-time estimates at sub-national scale.
 
+## Mental Health Treatment Gap Measurement In Low- And Middle-Income Countries
+
+- ID: `public-health/mental-health-treatment-gap-global`
+- Status: `scoped`
+- Domain: `public-health`
+- Region: `global`
+- Source: [problem-packs/public-health/mental-health-treatment-gap-global](../../problem-packs/public-health/mental-health-treatment-gap-global/problem.md)
+
+Build a verified workflow for measuring the mental health treatment gap — the proportion of people with mental disorders who receive no care — in low- and middle-income countries, linking prevalence surveys, health-system capacity data, and treatment-seeking behavior surveys.
+
 ## Child Stunting Hotspot Identification In Sub-Saharan Africa
 
 - ID: `public-health/stunting-sub-saharan-africa`