fix: repair rotted evidence-source URLs + scheduled link-rot CI

.github/workflows/verify-sources-scheduled.yml

@@ -0,0 +1,36 @@
+name: Scheduled Source Health Check
+
+# Periodically re-checks every evidence-record URL so link rot is caught
+# automatically instead of only when a pack is next edited. This is a health
+# check, not a PR gate: a failure means a source needs re-verification or
+# re-sourcing, surfaced via the Actions run and its failure notification.
+
+on:
+  schedule:
+    # 07:00 UTC every Monday
+    - cron: "0 7 * * 1"
+  workflow_dispatch:
+
+jobs:
+  verify-sources:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 10
+
+      - name: Setup Node
+        uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Verify evidence-source URLs
+        run: pnpm verify:sources

problem-packs/biodiversity/grassland-degradation-central-asia/evidence.json

@@ -2,22 +2,22 @@
   {
     "id": "fao-grassland-degradation-2023",
     "problem_id": "biodiversity/grassland-degradation-central-asia",
-    "claim": "The FAO Status of the World's Grasslands estimates that 20-30 percent of global grasslands are degraded, with Central Asian steppes and rangelands experiencing some of the highest degradation rates globally, driven primarily by overgrazing, conversion to cropland, and climate-change-induced aridification.",
+    "claim": "FAO's first Global Assessment of Soil Carbon in Grasslands (2023) finds grasslands hold about 20 percent of the world's soil organic carbon stocks and reports a negative carbon balance — declining stocks indicative of degradation pressure — in several regions including Africa south of the Equator, Central and South America, and East Asia.",
     "evidence_type": "primary-source",
     "source": {
-      "title": "Status of the World's Grasslands",
-      "url": "https://www.fao.org/publications/grasslands/2023"
+      "title": "Global assessment of soil carbon in grasslands",
+      "url": "https://openknowledge.fao.org/items/29349846-3fec-4623-b5bc-0984a860ec43"
     },
-    "source_date": "2023-01-01",
-    "access_date": "2026-06-07",
-    "method": "Reviewed FAO grassland assessment for degradation statistics, driver analysis, and regional comparisons. Focused on Central Asian country-specific data and methodology for classifying degradation.",
+    "source_date": "2023-02-01",
+    "access_date": "2026-06-18",
+    "method": "Re-sourced after the originally cited 'Status of the World's Grasslands' link rotted (HTTP 404) and could not be confirmed to exist under that title; replaced with FAO's verifiable 2023 grassland soil-carbon assessment and rewrote the claim to match what that report actually supports.",
     "limitations": [
-      "The 20-30 percent figure is a global estimate — Central Asian degradation rates may be higher or lower depending on methodology and country.",
-      "Grassland-degradation classification varies by methodology — remote-sensing, field survey, and expert assessment give different results.",
-      "The assessment does not provide sub-national degradation maps suitable for pastoral livelihood analysis.",
-      "Livelihood impact of grassland degradation is discussed qualitatively but not quantified with household-level data."
+      "This report measures soil organic carbon stocks and balance, not a direct percent-degraded statistic; the prior '20-30 percent degraded' figure was not supported by a verifiable FAO source and has been removed.",
+      "Regional carbon-balance findings do not single out Central Asia; a Central-Asia-specific degradation-extent figure still needs a dedicated source.",
+      "Soil-carbon balance is one proxy for degradation; it does not capture vegetation cover loss, biodiversity, or pastoral livelihood impacts.",
+      "Estimates are modeled from global datasets with country-level uncertainty."
     ],
-    "confidence": "high"
+    "confidence": "low"
   },
   {
     "id": "modis-ndvi-grassland-monitoring",

problem-packs/biodiversity/grassland-degradation-central-asia/evidence.md

@@ -6,9 +6,9 @@ The machine-readable ledger is `evidence.json`.
 
 ## Evidence Notes
 
-### FAO Grassland Degradation 2023
+### FAO Global Assessment of Soil Carbon in Grasslands (2023)
 
-Use this source for the global grassland degradation baseline and Central Asian context. The 20-30 percent figure is a global estimate — degradation rates vary by methodology and country. Do not use this as proof of degradation in any specific Central Asian sub-national area without satellite or field verification. The FAO assessment does not provide sub-national maps suitable for pastoral livelihood analysis.
+Re-sourced in June 2026: the original "Status of the World's Grasslands" citation rotted (HTTP 404) and could not be confirmed to exist under that title, and its "20-30 percent degraded" figure was unsupported. Replaced with FAO's verifiable 2023 grassland soil-carbon assessment, which supports that grasslands hold ~20 percent of global soil organic carbon and shows a negative carbon balance (declining stocks) in several regions. This is a soil-carbon proxy, not a percent-degraded statistic, and it does not single out Central Asia — a Central-Asia-specific degradation-extent figure still needs a dedicated source. Do not use this as proof of degradation in any specific Central Asian sub-national area without satellite or field verification.
 
 ### MODIS NDVI Grassland Monitoring
 

problem-packs/biodiversity/pollinator-decline-agriculture-global/evidence.json

@@ -58,25 +58,5 @@
       "Temporal resolution of land-cover maps (annual or multi-year composites) may miss seasonal habitat changes important for pollinator life cycles."
     ],
     "confidence": "medium"
-  },
-  {
-    "id": "pollinator-monitoring-gap-tropical",
-    "problem_id": "biodiversity/pollinator-decline-agriculture-global",
-    "claim": "Systematic wild pollinator population monitoring is almost entirely absent from Sub-Saharan Africa, South Asia, and Southeast Asia — the regions where pollinator-dependent crop production is growing fastest. Over 80 percent of pollinator studies globally come from Europe and North America.",
-    "evidence_type": "peer-reviewed-study",
-    "source": {
-      "title": "Global pollinator monitoring gaps and implications for food security",
-      "url": "https://www.sciencedirect.com/science/article/pii/S0169534722002345"
-    },
-    "source_date": "2022-11-01",
-    "access_date": "2026-06-07",
-    "method": "Reviewed systematic search of pollinator monitoring programs by region, study-design quality assessment, and geographic coverage analysis.",
-    "limitations": [
-      "The 80 percent figure is from the IPBES assessment; more recent studies suggest the imbalance persists but exact percentages may have shifted slightly.",
-      "Some LMIC countries have emerging pollinator monitoring programs not yet captured in global reviews.",
-      "Absence of monitoring data does not mean absence of pollinator decline — it means the decline, if occurring, is undetected.",
-      "Citizen-science programs (e.g., iNaturalist) provide some opportunistic LMIC pollinator occurrence data but with severe spatial and taxonomic bias."
-    ],
-    "confidence": "high"
   }
 ]

problem-packs/biodiversity/pollinator-decline-agriculture-global/evidence.md

@@ -18,9 +18,7 @@ Use MapSPAM for sub-national crop-pollination dependence mapping, with the cavea
 
 Use satellite land-cover data as an indirect proxy for pollinator habitat pressure, not as a direct measure of pollinator abundance or diversity. Habitat fragmentation, agricultural intensification, and landscape simplification are documented correlates of pollinator decline primarily from European and North American studies. The transferability to tropical agricultural landscapes is uncertain. Land-cover classification accuracy is lower in tropical and dryland agricultural mosaics. Always document that the proxy-to-pollinator inference chain has not been validated with ground monitoring in the analyzed region.
 
-### Pollinator Monitoring Gap Documentation
-
-Use this source to document the fundamental evidence constraint for this pack: systematic wild pollinator monitoring is almost entirely absent from the LMIC regions where this pack focuses. This means any pollination-service risk map for Sub-Saharan Africa, South Asia, or Southeast Asia must be understood as a proxy-based risk indicator, not a verified decline measurement. Document this gap explicitly in every output.
+> Note (June 2026): a separate "Pollinator Monitoring Gap" record was removed because its cited source URL had rotted (HTTP 404) and its claim duplicated the IPBES geographic-evidence-gap finding documented above. The constraint — systematic wild pollinator monitoring is largely absent from Sub-Saharan Africa, South Asia, and Southeast Asia — is captured by the IPBES record; treat any LMIC pollination-service risk map as a proxy indicator, not a verified decline measurement.
 
 ## Evidence Quality Rule
 

problem-packs/climate-adaptation/heat-early-warning-systems-global/evidence.json

@@ -5,19 +5,20 @@
     "claim": "WMO's State of Climate Services 2023 report on health estimates that fewer than 40 LMICs have operational heat-health early warning systems. Where systems exist, systematic evaluation of mortality-reduction effectiveness is rare — fewer than 10 countries globally have published heat-wave mortality time series with pre- and post-warning-system comparison.",
     "evidence_type": "primary-source",
     "source": {
-      "title": "WMO State of Climate Services 2023: Health",
-      "url": "https://public.wmo.int/en/resources/library/state-of-climate-services-2023-health"
+      "title": "2023 State of Climate Services: Health (WMO-No. 1335)",
+      "url": "https://library.wmo.int/records/item/68500-2023-state-of-climate-services-health"
     },
     "source_date": "2023-11-02",
-    "access_date": "2026-06-07",
+    "access_date": "2026-06-18",
     "method": "Reviewed WMO report for heat-health warning system country coverage, operational status classification, and effectiveness evaluation assessment. Focused on the gap between system existence and verified effectiveness.",
     "limitations": [
       "The 'fewer than 40' figure is a binary existence indicator — it does not capture whether warnings reach vulnerable populations, trigger protective actions, or reduce mortality.",
       "Effectiveness evaluation data is sparse — the report itself notes the absence of systematic evaluation in most countries.",
       "System coverage does not guarantee effective response — the gap between warning issuance and protective action is not captured by the existence indicator.",
-      "Country-level coverage data may not capture sub-national variation — some countries have national systems that are operational only in major cities."
+      "Country-level coverage data may not capture sub-national variation — some countries have national systems that are operational only in major cities.",
+      "Source URL was repaired in June 2026 after the original link rotted; the specific country counts cited above should be reconfirmed against the WMO-No. 1335 PDF on next review."
     ],
-    "confidence": "high"
+    "confidence": "medium"
   },
   {
     "id": "ahmedabad-heat-action-plan",