cschanhniem · Jun 28, 2026 · Jun 28, 2026 · Jun 28, 2026 · Jun 28, 2026 · Jun 28, 2026
diff --git a/agent-radar.json b/agent-radar.json
@@ -52,6 +52,22 @@
         "tasks_file": "problem-packs/public-health/birth-registration-access-global/tasks.json",
         "why_pick_now": "Completing this scoped task opens 5 follow-on tasks across 4 additional roles."
       },
+      {
+        "pack_id": "public-health/ncd-risk-factor-surveillance-global",
+        "pack_title": "Non-Communicable Disease Risk Factor Surveillance Gaps In Low-Income Countries",
+        "task_id": "source-inventory",
+        "title": "Inventory data sources for public-health/ncd-risk-factor-surveillance-global",
+        "owner_role": "literature-scout",
+        "reviewer_needed": "domain-reviewer",
+        "safety_risk": "low",
+        "done_condition": "At least four sources classified as usable, limited, or rejected with explicit reasons.",
+        "evidence_count": 5,
+        "downstream_tasks_unlocked": 4,
+        "downstream_high_risk_tasks": 2,
+        "problem_file": "problem-packs/public-health/ncd-risk-factor-surveillance-global/problem.md",
+        "tasks_file": "problem-packs/public-health/ncd-risk-factor-surveillance-global/tasks.json",
+        "why_pick_now": "Completing this scoped task opens 4 follow-on tasks across 4 additional roles."
+      },
       {
         "pack_id": "public-health/oral-health-access-global",
         "pack_title": "Oral Health Service Access And Dental Workforce Gaps In Low-Income Countries",
@@ -100,22 +116,6 @@
         "tasks_file": "problem-packs/education/skills-training-youth-employment-global/tasks.json",
         "why_pick_now": "Completing this scoped task opens 4 follow-on tasks across 4 additional roles."
       },
-      {
-        "pack_id": "public-health/ncd-risk-factor-surveillance-global",
-        "pack_title": "Non-Communicable Disease Risk Factor Surveillance Gaps In Low-Income Countries",
-        "task_id": "source-inventory",
-        "title": "Inventory data sources for public-health/ncd-risk-factor-surveillance-global",
-        "owner_role": "literature-scout",
-        "reviewer_needed": "domain-reviewer",
-        "safety_risk": "low",
-        "done_condition": "At least four sources classified as usable, limited, or rejected with explicit reasons.",
-        "evidence_count": 1,
-        "downstream_tasks_unlocked": 4,
-        "downstream_high_risk_tasks": 2,
-        "problem_file": "problem-packs/public-health/ncd-risk-factor-surveillance-global/problem.md",
-        "tasks_file": "problem-packs/public-health/ncd-risk-factor-surveillance-global/tasks.json",
-        "why_pick_now": "Completing this scoped task opens 4 follow-on tasks across 4 additional roles."
-      },
       {
         "pack_id": "climate-health/malaria-early-warning-africa",
         "pack_title": "Malaria Early Warning Signals In Sub-Saharan Africa",

diff --git a/docs/wiki/Agent-Radar.md b/docs/wiki/Agent-Radar.md
@@ -37,7 +37,19 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris
 - Why pick now: Completing this scoped task opens 5 follow-on tasks across 4 additional roles.
 - Done condition: At least six candidate sources are classified as usable, limited, or rejected with explicit reasons covering measure family, timing, geographic grain, and health-linkage relevance.
 
-### 2. Oral Health Service Access And Dental Workforce Gaps In Low-Income Countries
+### 2. Non-Communicable Disease Risk Factor Surveillance Gaps In Low-Income Countries
+
+- Pack: [`public-health/ncd-risk-factor-surveillance-global`](../../problem-packs/public-health/ncd-risk-factor-surveillance-global/problem.md)
+- Task: `source-inventory` — Inventory data sources for public-health/ncd-risk-factor-surveillance-global
+- Risk: `low`
+- Reviewer needed: `domain-reviewer`
+- Existing evidence records: 5
+- Downstream tasks unlocked: 4
+- Downstream high-risk tasks: 2
+- Why pick now: Completing this scoped task opens 4 follow-on tasks across 4 additional roles.
+- Done condition: At least four sources classified as usable, limited, or rejected with explicit reasons.
+
+### 3. Oral Health Service Access And Dental Workforce Gaps In Low-Income Countries
 
 - Pack: [`public-health/oral-health-access-global`](../../problem-packs/public-health/oral-health-access-global/problem.md)
 - Task: `source-inventory` — Inventory data sources for public-health/oral-health-access-global
@@ -49,7 +61,7 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris
 - Why pick now: Completing this scoped task opens 4 follow-on tasks across 4 additional roles.
 - Done condition: At least four sources classified as usable, limited, or rejected with explicit reasons.
 
-### 3. Digital Divide Measurement And School Internet Connectivity In Low-Income Countries
+### 4. Digital Divide Measurement And School Internet Connectivity In Low-Income Countries
 
 - Pack: [`education/digital-divide-school-access-global`](../../problem-packs/education/digital-divide-school-access-global/problem.md)
 - Task: `source-inventory` — Inventory school connectivity, mobile-network, and digital-learning data sources for low-income countries
@@ -61,7 +73,7 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris
 - Why pick now: Completing this scoped task opens 4 follow-on tasks across 4 additional roles.
 - Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons covering connectivity-classification methodology, data currency, and cross-validation status.
 
-### 4. Youth Skills Training And Employment Outcome Gaps In Low-Income Countries
+### 5. Youth Skills Training And Employment Outcome Gaps In Low-Income Countries
 
 - Pack: [`education/skills-training-youth-employment-global`](../../problem-packs/education/skills-training-youth-employment-global/problem.md)
 - Task: `source-inventory` — Inventory training-provider, employment-outcome, labor-market, and skills-mismatch data sources for LMICs
@@ -73,18 +85,6 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris
 - Why pick now: Completing this scoped task opens 4 follow-on tasks across 4 additional roles.
 - Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons covering provider-data fragmentation, outcome-tracking rates, and informal-sector measurement gaps.
 
-### 5. Non-Communicable Disease Risk Factor Surveillance Gaps In Low-Income Countries
-
-- Pack: [`public-health/ncd-risk-factor-surveillance-global`](../../problem-packs/public-health/ncd-risk-factor-surveillance-global/problem.md)
-- Task: `source-inventory` — Inventory data sources for public-health/ncd-risk-factor-surveillance-global
-- Risk: `low`
-- Reviewer needed: `domain-reviewer`
-- Existing evidence records: 1
-- Downstream tasks unlocked: 4
-- Downstream high-risk tasks: 2
-- Why pick now: Completing this scoped task opens 4 follow-on tasks across 4 additional roles.
-- Done condition: At least four sources classified as usable, limited, or rejected with explicit reasons.
-
 ### 6. Malaria Early Warning Signals In Sub-Saharan Africa
 
 - Pack: [`climate-health/malaria-early-warning-africa`](../../problem-packs/climate-health/malaria-early-warning-africa/problem.md)

diff --git a/problem-packs/climate-health/air-pollution-source-attribution-global/evidence.json b/problem-packs/climate-health/air-pollution-source-attribution-global/evidence.json
@@ -25,17 +25,18 @@
     "claim": "Source-apportionment methods — chemical tracer analysis, positive matrix factorization, and chemical transport modeling — give different source-contribution estimates for the same city, with methodology choice explaining 10-30 percent of the variance in attribution percentages across studies.",
     "evidence_type": "peer-reviewed-study",
     "source": {
-      "title": "Intercomparison of source apportionment methods for PM2.5 in megacities",
-      "url": "https://www.atmos-chem-phys.net/21/12345/2021/"
+      "title": "Atmospheric Chemistry and Physics — Copernicus Publications",
+      "url": "https://acp.copernicus.org/"
     },
     "source_date": "2021-09-15",
-    "access_date": "2026-06-07",
+    "access_date": "2026-06-28",
     "method": "Reviewed intercomparison studies of source-apportionment methods, methodology-specific bias analysis, and recommendations for LMIC application.",
     "limitations": [
       "Intercomparison studies are conducted at well-characterized sites — methodology performance at data-sparse LMIC sites is less studied.",
       "The 10-30 percent variance figure depends on which methods are compared and which source categories are examined.",
       "Chemical transport models require emission inventories that have higher uncertainty in LMIC cities with informal economic activity.",
-      "Receptor models (PMF) require speciated PM2.5 monitoring data that is available at very few LMIC sites."
+      "Receptor models (PMF) require speciated PM2.5 monitoring data that is available at very few LMIC sites.",
+      "URL changed from a placeholder article page (atmos-chem-phys.net/21/12345/2021, unreachable) to the ACP journal homepage; the specific article should be re-identified by DOI."
     ],
     "confidence": "high"
   },

diff --git a/problem-packs/disaster-resilience/volcanic-ash-aviation-global/evidence.json b/problem-packs/disaster-resilience/volcanic-ash-aviation-global/evidence.json
@@ -45,17 +45,18 @@
     "claim": "Ash-transport models (HYSPLIT, NAME, FLEXPART) forecast volcanic ash plume dispersion but have substantial uncertainty from eruption-source parameters — ash column height, mass eruption rate, and particle-size distribution are rarely measured in real time and must be estimated from limited observations.",
     "evidence_type": "peer-reviewed-study",
     "source": {
-      "title": "Volcanic ash transport model uncertainty and eruption source parameter sensitivity",
-      "url": "https://www.atmos-chem-phys.net/22/12345/2022/"
+      "title": "Atmospheric Chemistry and Physics — Copernicus Publications",
+      "url": "https://acp.copernicus.org/"
     },
     "source_date": "2022-08-15",
-    "access_date": "2026-06-07",
+    "access_date": "2026-06-28",
     "method": "Reviewed ash-transport model intercomparison studies, eruption-source parameter sensitivity analysis, and forecast validation against satellite observations.",
     "limitations": [
       "Model validation is limited to well-observed eruptions — forecast accuracy for poorly monitored eruptions (remote volcanoes, submarine eruptions) is largely unknown.",
       "Eruption-source parameter uncertainty propagates non-linearly through transport models — small errors in ash height can produce large errors in downwind concentration forecasts.",
       "Model validation uses satellite ash detection as ground truth, but satellite detection itself has 20-40 percent false-positive rates — the validation chain is circular.",
-      "Ensemble modeling approaches can quantify forecast uncertainty but require multiple model runs that are computationally expensive for operational use."
+      "Ensemble modeling approaches can quantify forecast uncertainty but require multiple model runs that are computationally expensive for operational use.",
+      "URL changed from a placeholder article page (atmos-chem-phys.net/22/12345/2022, unreachable) to the ACP journal homepage; the specific article should be re-identified by DOI."
     ],
     "confidence": "high"
   },

diff --git a/problem-packs/public-health/ncd-risk-factor-surveillance-global/claims.json b/problem-packs/public-health/ncd-risk-factor-surveillance-global/claims.json
@@ -0,0 +1,36 @@
+[
+  {
+    "id": "ncd-surveillance-data-scarcity-lmics",
+    "problem_id": "public-health/ncd-risk-factor-surveillance-global",
+    "claim": "Available NCD risk factor surveillance data sources are sufficient for country-level gap identification in low-income countries (which countries have STEPS surveys, when they were last conducted, and what modeled estimates exist), but no source provides sub-national NCD risk factor data for any low-income country. Sub-national risk factor gap mapping is not currently feasible with public data.",
+    "domain": ["public-health"],
+    "status": "dry-lab-verified",
+    "evidence": [
+      "who-steps-surveillance-2023",
+      "ihme-gbd-2021",
+      "who-gho-ncd-indicators",
+      "who-ncd-surveillance-data-portal",
+      "dhis2-ncd-modules"
+    ],
+    "failure_modes": [
+      "A reader treats modeled GBD estimates as equivalent to direct surveillance, masking the fact that most LICs have no recent STEPS data.",
+      "A reader assumes country-level surveillance presence means sub-national coverage, when STEPS sub-national sampling varies by country.",
+      "A reader treats DHIS2 NCD module availability as evidence of implementation, when adoption is the bottleneck and most LIC deployments focus on infectious disease reporting.",
+      "A reader uses heterogeneous GHO indicators as if they were standardized across countries, when underlying survey methodology and year vary significantly.",
+      "The claim about fewer than 40 percent repeat STEPS surveys is an inference from the existing evidence record and WHO portal content, not a single WHO-stated figure, and could be wrong if the denominator or time window differs."
+    ],
+    "kill_condition": "A reviewer identifies a public dataset that provides sub-national NCD risk factor data (measured or self-reported) for at least one low-income country, or finds that WHO has published a systematic assessment of STEPS repeat-survey frequency with a figure different from the 40 percent inference.",
+    "review_required": ["domain-reviewer"],
+    "safety_level": "low",
+    "submitter": "codex-agent",
+    "created_date": "2026-06-28",
+    "last_updated": "2026-06-28",
+    "confidence": "medium",
+    "limitations": [
+      "The claim is about data availability, not about NCD outcomes or risk factor prevalence.",
+      "Country-level STEPS data is self-reported by health ministries and survey timing is not synchronized.",
+      "The inference about 40 percent repeat-survey frequency depends on the existing evidence record classification, which should be verified against WHO STEPS country reports.",
+      "DHIS2 NCD module implementation status is inferred from documentation, not from a systematic survey of all LIC deployments."
+    ]
+  }
+]
diff --git a/problem-packs/public-health/ncd-risk-factor-surveillance-global/datasets.md b/problem-packs/public-health/ncd-risk-factor-surveillance-global/datasets.md
@@ -1,15 +1,27 @@
 # Dataset Inventory
 
-| Source                       | Grain        | Status | Use           |
-| ---------------------------- | ------------ | ------ | ------------- |
-| Primary survey data          | Sub-national | Usable | Core analysis |
-| Satellite/environmental data | Variable     | Usable | Risk context  |
-| Population data              | 100m         | Usable | Denominator   |
+## Candidate Sources
 
-## Required Properties
+| Source                             | Grain                 | Status   | Use                                               | Reason for classification                                                                                                                                                                                                    |
+| ---------------------------------- | --------------------- | -------- | ------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| WHO STEPS Surveillance Manual      | Country               | Usable   | Risk factor survey coverage and frequency         | Standardized NCD risk factor survey protocol used in 120+ countries. Country-level; sub-national data available only for selected countries where STEPS was designed with sub-national sampling.                             |
+| IHME Global Burden of Disease 2021 | Country               | Usable   | Modeled risk factor estimates, trend comparison   | Modeled estimates for 204 countries from 1990-2021. Sub-national for only a subset. Modeled not measured; useful for trend comparison and gap identification, not as a substitute for surveillance.                          |
+| WHO GHO NCD Indicators             | Country               | Usable   | Cross-country risk factor comparison              | Country-level NCD risk factor indicators for most UN member states. Heterogeneous underlying sources. No sub-national data for any LIC. Useful for cross-country comparison, not for sub-national gap mapping.               |
+| WHO NCD Surveillance Data Portal   | Country               | Usable   | Survey availability and capacity assessment       | Lists STEPS survey availability and country capacity. Does not provide downloadable sub-national data. Useful for identifying which countries have surveillance and which have gaps.                                         |
+| DHIS2 Health Information System    | Facility/Sub-district | Limited  | Facility-level NCD service data, not risk factors | Used by 80+ countries but NCD module adoption is limited in LICs. Captures service data, not population risk factor surveys. Complementary to STEPS, not a replacement. Useful where implemented; not universally available. |
+| DHS / MICS NCD modules             | Sub-national          | Rejected | Insufficient coverage                             | DHS and MICS do not include comprehensive NCD risk factor examination modules in most LICs. Where NCD questions exist, they are limited to self-reported hypertension and diabetes diagnosis, with no biomarker measurement. |
 
-Source methodology, data year, geographic grain, license.
+## Required Dataset Properties
+
+- Date range.
+- Geographic grain.
+- Risk factor definition (measured vs self-reported).
+- Survey methodology (STEPS, national health survey, modeled estimate).
+- Missingness.
+- License or reuse permission.
+- Denominator source.
+- Known changes in survey methodology over time.
 
 ## Rejection Rule
 
-Rejected if methodology, grain, date, or license unverifiable.
+A dataset is rejected for canonical modeling if grain, date range, license, or method cannot be verified. Rejected datasets may still be listed as context.