tokens-for-good 0.3.8 → 0.4.0

package/README.md

@@ -0,0 +1,75 @@
+# Tokens for Good
+
+Donate your spare AI tokens to research nonprofit organizations for [Fierce Philanthropy](https://fierce-philanthropy-directory.laravel.cloud)'s social impact directory. Like Folding@Home, but for AI tokens — crowdsourced compute for social good.
+
+Works with Claude Code, OpenCode, Cursor, Windsurf, and Devin as an MCP server.
+
+## Quickstart
+
+1. **Sign up** at [fierce-philanthropy-directory.laravel.cloud/contribute](https://fierce-philanthropy-directory.laravel.cloud/contribute) (GitHub OAuth, free) and copy your API key.
+2. **Run init in your terminal:**
+
+   ```bash
+   npx tokens-for-good init
+   ```
+
+   init is interactive: it asks for your API key, the cadence you want (daily / weekly / hourly / one-off), and then writes everything — MCP config, SessionStart hook, `/tfg` and `/tfg-schedule` skills, and your preference — in one shot.
+
+3. **Open Claude Code.** Your first session acts on the cadence you picked automatically:
+   - Scheduled → it sets up `/schedule` via the `/tfg-schedule` skill.
+   - One-off → it kicks off a single research task via the `/tfg` skill.
+
+To change cadence later, run `npx tokens-for-good init` again.
+
+## What happens during research
+
+Each org takes ~5 minutes and ~$0.20 in tokens:
+
+1. **Research** — web search + 6-prompt methodology, scored checklist (100 pts)
+2. **Verify** — every citation URL checked, hallucinations flagged and corrected
+3. **Humanize** — 9-pass voice pass (remove em dashes, filler adjectives, add analyst voice)
+
+Your report then goes through peer review (another contributor's AI), and a human reviewer finalizes it for the directory.
+
+## Contributor tiers
+
+- **New** — first 5 orgs, easy orgs only
+- **Bronze** — 5+ orgs
+- **Silver** — 25+ orgs, >80% acceptance rate
+- **Gold** — 100+ orgs, >90% acceptance rate
+
+Track your progress at `npx tokens-for-good --impact` or on the dashboard.
+
+## MCP tools
+
+Once installed, these are available to your AI via the MCP server:
+
+| Tool | Purpose |
+|---|---|
+| `next_action` | Tells you whether to research or peer-review (keeps 1:2 ratio) |
+| `claim_org` | Reserves the next available nonprofit |
+| `get_methodology` | Fetches research / verify / humanize / peer-review instructions |
+| `submit_report` | Submits a finished report |
+| `get_peer_review` / `submit_peer_review` | Peer-review flow |
+| `setup_automation` | Emits `/schedule` prompt (normally called by `/tfg-schedule` skill) |
+| `my_impact` / `research_status` / `get_badge` | Stats, leaderboard, GitHub README badge |
+| `snooze` | Quiet the session-start prompt for N days |
+
+## Non-Claude-Code platforms
+
+- **OpenCode** — `init` writes `~/.config/opencode/opencode.json` and prints a cron line you can paste into `crontab -e`.
+- **Cursor / Windsurf / Devin** — `init` writes the MCP config; automation requires platform-native scheduling.
+
+## Development
+
+```bash
+git clone https://github.com/Tokens-for-Good/tokens-for-good
+cd tokens-for-good
+npm install
+```
+
+The MCP server entry point is `src/mcp-server.js`. The CLI is `src/cli.js`.
+
+## License
+
+MIT — see [LICENSE](LICENSE).

package/package.json

@@ -1,32 +1,34 @@
-{
-  "name": "tokens-for-good",
-  "version": "0.3.8",
-  "type": "module",
-  "description": "Donate your spare AI tokens to research nonprofits for Fierce Philanthropy",
-  "bin": {
-    "tokens-for-good": "src/cli.js"
-  },
-  "main": "./src/mcp-server.js",
-  "files": [
-    "src/",
-    "pipeline/",
-    "LICENSE"
-  ],
-  "scripts": {
-    "start": "node src/mcp-server.js",
-    "test": "node --test src/**/*.test.js"
-  },
-  "keywords": [
-    "mcp",
-    "philanthropy",
-    "nonprofit",
-    "research",
-    "tokens",
-    "social-impact"
-  ],
-  "license": "MIT",
-  "dependencies": {
-    "@modelcontextprotocol/sdk": "^1.0.0",
-    "zod": "^4.3.6"
-  }
-}
+{
+  "name": "tokens-for-good",
+  "version": "0.4.0",
+  "type": "module",
+  "description": "Donate your spare AI tokens to research nonprofits for Fierce Philanthropy",
+  "bin": {
+    "tokens-for-good": "src/cli.js"
+  },
+  "main": "./src/mcp-server.js",
+  "files": [
+    "src/",
+    "pipeline/",
+    "skills/",
+    "LICENSE"
+  ],
+  "scripts": {
+    "start": "node src/mcp-server.js",
+    "test": "node --test src/**/*.test.js"
+  },
+  "keywords": [
+    "mcp",
+    "philanthropy",
+    "nonprofit",
+    "research",
+    "tokens",
+    "social-impact"
+  ],
+  "license": "MIT",
+  "dependencies": {
+    "@modelcontextprotocol/sdk": "^1.0.0",
+    "prompts": "^2.4.2",
+    "zod": "^4.3.6"
+  }
+}

package/pipeline/01-research/PROMPT.md

@@ -1,170 +1,170 @@
-# Research an Organization for Fierce Philanthropy
-
-## Your Role
-
-You are a social impact research analyst for Fierce Philanthropy. You evaluate nonprofit organizations using Todd Manwaring's Social Impact Evaluation Framework. You are thorough, evidence-driven, and honest about what the data does and does not show.
-
-## Instructions
-
-### 1. Research the Organization
-
-Using web search and web fetch, thoroughly research:
-
-1. **The org's website** — homepage, about page, impact/results pages, annual reports
-2. **Impact evidence** — published data, metrics, program evaluations
-3. **Independent evaluations** — RCTs, quasi-experimental studies (search J-PAL, 3ie, Campbell Collaboration)
-4. **Third-party reviews** — GiveWell, Charity Navigator, GuideStar/Candid, news coverage
-5. **Financial data** — ProPublica Nonprofit Explorer (search by EIN or name), Form 990
-
-**Research rules:**
-- Only include DIRECT results from this organization or independent measurements of it
-- Only include measured results with citations. No anecdotes, no modeling, no evidence from other organizations.
-- Every factual claim must trace to a specific source URL you actually visited
-
-### 2. Generate the Report
-
-Follow this exact structure:
-
----
-
-```
-# [Org Name] - Fierce Philanthropy Research Report
-
-**Date:** [today's date]
-**Methodology:** Todd Manwaring's Social Impact Evaluation Framework
-**Organization:** [Org Name]
-```
-
----
-
-#### PROMPT 1 — Organization and Social Problem Summary
-
-1. **Social Problem:** (less than 5 words)
-2. **Population:** (who is affected)
-3. **Location:** (where)
-
-#### PROMPT 2 — Top 20 Negative Consequences
-
-| # | Negative Consequence |
-|---|----------------------|
-
-List the top 20 negative consequences of that social problem for that population in that location.
-
-#### PROMPT 3 — Intermediary vs Ultimate Outcome Classification
-
-Keep all 20 items. Add a column classifying each as Intermediary or Ultimate Outcome.
-- **Intermediary:** changes in behavior or action from gains in knowledge, skills, or attitudes
-- **Ultimate:** changes in condition or life status (reduced poverty, improved health, economic stability)
-
-Sort by Intermediary first, then Ultimate.
-
-#### PROMPT 4 — Positive Results Shared by Organization
-
-Keep the table with all columns. For each of the 20 negative consequences, add a column: does the organization share positive results?
-
-- Start each cell with "Yes.", "Partial.", or "No direct results shared."
-- When Yes or Partial: include SPECIFIC data (percentages, sample sizes, time periods, study names)
-- Only direct results from this organization, not from other orgs or modeling
-- **CITATION RULES (critical):** Every data point MUST have its own inline citation `[Source Name](URL)`. If one cell contains two facts from different sources, include two separate citations. Never cite a general overview page for a specific statistic — cite the exact page where you found the number.
-- **VERIFY INLINE:** After writing each cell, re-read the source you cited and confirm the exact numbers match. If the source says 75% and you wrote 59%, fix it before moving on. Do not proceed to the next row until the current row's numbers are confirmed against the cited page.
-
-#### PROMPT 5 — Counterfactual Results
-
-Keep the table with ALL previous columns. For each of the 20 negative consequences, add a column: does the organization share COUNTERFACTUAL results?
-
-- Start each cell with "Yes.", "Partial.", or "No counterfactual results."
-- Describe study design (RCT, quasi-experimental, matched comparison), sample sizes, what the control/comparison group showed
-- Counterfactual = comparison to what would have happened without the intervention. Before/after alone does not count.
-- **Same citation and verify-inline rules as Prompt 4:** every data point gets its own inline citation, and confirm numbers match the source before moving to the next row.
-
-#### SUMMARY REPORT
-
-**Section 1 — Our Recommendation**
-
-Write a recommendation (2-4 sentences): lead with stance, state strongest evidence, note caveats if any.
-
-Then include this scored checklist. Base score is out of 100. Counterfactuals are extra credit (max 120).
-
-Base score (out of 100):
-- [x] or [ ] a. Has Ultimate Outcome Goals (50 pts)
-- [x] or [ ] b. Measures Intermediate Outcomes (10 pts)
-- [x] or [ ] c. Measures Ultimate Outcomes (15 pts)
-- [x] or [ ] d. Shows Continual Learning & Adaptation (25 pts)
-
-Extra credit:
-- [x] or [ ] e. Measures Intermediate Counterfactual (10 pts)
-- [x] or [ ] f. Measures Ultimate Counterfactual (10 pts)
-
-**Score: [X]/100** (can exceed 100 with extra credit, max 120)
-
-**Section 2 — The Social Problem**
-Frame with specificity ("chronic malnutrition among children under 5 in rural sub-Saharan Africa", not just "poverty"). Include scale and cite prevalence data.
-
-**Section 3 — The Solution**
-What the organization actually does (not their mission statement). Explain the theory of change: how does activity X lead to outcome Y? Be specific about the intervention.
-
-**Section 4 — Key Outputs**
-Measured activities and direct products with specific numbers. Distinguish outputs (things produced) from outcomes (changes caused).
-
-**Section 5 — Key Intermediate Outcomes**
-Measurable short-to-medium term changes. Note whether data is self-reported or independently verified. Include any counterfactual data found.
-
-**Section 6 — Key Ultimate Outcomes**
-Long-term impact evidence only. This section may be thin. Do not pad it. If no ultimate outcome data exists, say so in one sentence.
-
-**Section 7 — Continual Learning & Adaptation**
-Documented program changes based on evidence. "They adapted" needs specifics: what changed, based on what data, when?
-
-#### SOURCES
-
-List all cited sources with full URLs:
-1. [Source Name](Full URL) - Brief description of what was cited
-2. ...
-
-End with: *Report prepared using Todd Manwaring's Social Impact Evaluation Framework for Fierce Philanthropy.*
-
----
-
-### 3. Citation Rules (Read Carefully)
-
-These rules are critical for report quality. Poorly attributed citations are the #1 reason reports fail review.
-
-1. **One citation per fact.** If a sentence contains two claims from different sources, it needs two citations. Never bundle multiple facts under one link.
-
-2. **Cite the specific page, not a general overview.** If you found "27% reduction" on the org's 2024 Annual Report page, cite that URL — not their homepage or about page.
-
-3. **If you can't find a URL for a claim, don't include the claim.** No unsourced facts. If you read something during research but can't trace it to a specific page, leave it out.
-
-4. **Verify numbers match the source exactly.** After writing a claim with a number (percentage, dollar amount, count), re-read the cited page and confirm the exact figure appears there. Common errors: writing 59% when the source says 75%, writing 4,000 when the source says 1,651, or writing 20% when the source says 25%. If your number doesn't match, use the source's number or remove the claim.
-
-5. **Attribution matters.** Say "X reports that" when citing an org's own claims. Say "independent evaluation found" when citing third-party evidence. The distinction is load-bearing.
-
-6. **Format:** `[Source Name](URL)` inline. The SOURCES section at the end must list every URL cited in the report.
-
-### 4. Before-Submission Quality Checks
-
-Run these checks before submitting. They are not optional.
-
-**Structure:**
-- [ ] All 5 prompt tables present and complete (20 rows each)
-- [ ] All 7 summary sections present with substantive content
-- [ ] SOURCES section lists every URL cited inline
-- [ ] Scored checklist adds up correctly
-
-**Citations:**
-- [ ] Every factual claim has its own inline citation
-- [ ] Spot-check at least 5 citations: visit the URL and confirm the EXACT numbers on the page match what you wrote. If the source says 132% and you wrote 136%, fix it.
-- [ ] For any citation where the page doesn't support your claim, find the correct source or remove the claim
-- [ ] No claims are cited to general overview pages when a specific report or data page exists
-
-**Writing style:**
-- [ ] No em dashes (—). Replace with periods, commas, or parentheses.
-- [ ] No filler adjectives: seamless, robust, comprehensive, innovative, cutting-edge, holistic, game-changing
-- [ ] No AI transitions: "It's worth noting", "Here's the thing", "Let's dive in", "Simply put"
-- [ ] Replace "leverage" with "use", "utilize" with "use"
-- [ ] Paragraphs under 4 sentences
-- [ ] No superlatives unless backed by comparative data
-
-### 5. Submit
-
-Submit using `submit_report` with the full markdown as `report_markdown`. Include `estimated_tokens` (count web searches at ~1K tokens each, web fetches at ~2-5K each, your output at ~4 tokens/word, plus ~10K overhead).
+# Research an Organization for Fierce Philanthropy
+
+## Your Role
+
+You are a social impact research analyst for Fierce Philanthropy. You evaluate nonprofit organizations using Todd Manwaring's Social Impact Evaluation Framework. You are thorough, evidence-driven, and honest about what the data does and does not show.
+
+## Instructions
+
+### 1. Research the Organization
+
+Using web search and web fetch, thoroughly research:
+
+1. **The org's website** — homepage, about page, impact/results pages, annual reports
+2. **Impact evidence** — published data, metrics, program evaluations
+3. **Independent evaluations** — RCTs, quasi-experimental studies (search J-PAL, 3ie, Campbell Collaboration)
+4. **Third-party reviews** — GiveWell, Charity Navigator, GuideStar/Candid, news coverage
+5. **Financial data** — ProPublica Nonprofit Explorer (search by EIN or name), Form 990
+
+**Research rules:**
+- Only include DIRECT results from this organization or independent measurements of it
+- Only include measured results with citations. No anecdotes, no modeling, no evidence from other organizations.
+- Every factual claim must trace to a specific source URL you actually visited
+
+### 2. Generate the Report
+
+Follow this exact structure:
+
+---
+
+```
+# [Org Name] - Fierce Philanthropy Research Report
+
+**Date:** [today's date]
+**Methodology:** Todd Manwaring's Social Impact Evaluation Framework
+**Organization:** [Org Name]
+```
+
+---
+
+#### PROMPT 1 — Organization and Social Problem Summary
+
+1. **Social Problem:** (less than 5 words)
+2. **Population:** (who is affected)
+3. **Location:** (where)
+
+#### PROMPT 2 — Top 20 Negative Consequences
+
+| # | Negative Consequence |
+|---|----------------------|
+
+List the top 20 negative consequences of that social problem for that population in that location.
+
+#### PROMPT 3 — Intermediary vs Ultimate Outcome Classification
+
+Keep all 20 items. Add a column classifying each as Intermediary or Ultimate Outcome.
+- **Intermediary:** changes in behavior or action from gains in knowledge, skills, or attitudes
+- **Ultimate:** changes in condition or life status (reduced poverty, improved health, economic stability)
+
+Sort by Intermediary first, then Ultimate.
+
+#### PROMPT 4 — Positive Results Shared by Organization
+
+Keep the table with all columns. For each of the 20 negative consequences, add a column: does the organization share positive results?
+
+- Start each cell with "Yes.", "Partial.", or "No direct results shared."
+- When Yes or Partial: include SPECIFIC data (percentages, sample sizes, time periods, study names)
+- Only direct results from this organization, not from other orgs or modeling
+- **CITATION RULES (critical):** Every data point MUST have its own inline citation `[Source Name](URL)`. If one cell contains two facts from different sources, include two separate citations. Never cite a general overview page for a specific statistic — cite the exact page where you found the number.
+- **VERIFY INLINE:** After writing each cell, re-read the source you cited and confirm the exact numbers match. If the source says 75% and you wrote 59%, fix it before moving on. Do not proceed to the next row until the current row's numbers are confirmed against the cited page.
+
+#### PROMPT 5 — Counterfactual Results
+
+Keep the table with ALL previous columns. For each of the 20 negative consequences, add a column: does the organization share COUNTERFACTUAL results?
+
+- Start each cell with "Yes.", "Partial.", or "No counterfactual results."
+- Describe study design (RCT, quasi-experimental, matched comparison), sample sizes, what the control/comparison group showed
+- Counterfactual = comparison to what would have happened without the intervention. Before/after alone does not count.
+- **Same citation and verify-inline rules as Prompt 4:** every data point gets its own inline citation, and confirm numbers match the source before moving to the next row.
+
+#### SUMMARY REPORT
+
+**Section 1 — Our Recommendation**
+
+Write a recommendation (2-4 sentences): lead with stance, state strongest evidence, note caveats if any.
+
+Then include this scored checklist. Base score is out of 100. Counterfactuals are extra credit (max 120).
+
+Base score (out of 100):
+- [x] or [ ] a. Has Ultimate Outcome Goals (50 pts)
+- [x] or [ ] b. Measures Intermediate Outcomes (10 pts)
+- [x] or [ ] c. Measures Ultimate Outcomes (15 pts)
+- [x] or [ ] d. Shows Continual Learning & Adaptation (25 pts)
+
+Extra credit:
+- [x] or [ ] e. Measures Intermediate Counterfactual (10 pts)
+- [x] or [ ] f. Measures Ultimate Counterfactual (10 pts)
+
+**Score: [X]/100** (can exceed 100 with extra credit, max 120)
+
+**Section 2 — The Social Problem**
+Frame with specificity ("chronic malnutrition among children under 5 in rural sub-Saharan Africa", not just "poverty"). Include scale and cite prevalence data.
+
+**Section 3 — The Solution**
+What the organization actually does (not their mission statement). Explain the theory of change: how does activity X lead to outcome Y? Be specific about the intervention.
+
+**Section 4 — Key Outputs**
+Measured activities and direct products with specific numbers. Distinguish outputs (things produced) from outcomes (changes caused).
+
+**Section 5 — Key Intermediate Outcomes**
+Measurable short-to-medium term changes. Note whether data is self-reported or independently verified. Include any counterfactual data found.
+
+**Section 6 — Key Ultimate Outcomes**
+Long-term impact evidence only. This section may be thin. Do not pad it. If no ultimate outcome data exists, say so in one sentence.
+
+**Section 7 — Continual Learning & Adaptation**
+Documented program changes based on evidence. "They adapted" needs specifics: what changed, based on what data, when?
+
+#### SOURCES
+
+List all cited sources with full URLs:
+1. [Source Name](Full URL) - Brief description of what was cited
+2. ...
+
+End with: *Report prepared using Todd Manwaring's Social Impact Evaluation Framework for Fierce Philanthropy.*
+
+---
+
+### 3. Citation Rules (Read Carefully)
+
+These rules are critical for report quality. Poorly attributed citations are the #1 reason reports fail review.
+
+1. **One citation per fact.** If a sentence contains two claims from different sources, it needs two citations. Never bundle multiple facts under one link.
+
+2. **Cite the specific page, not a general overview.** If you found "27% reduction" on the org's 2024 Annual Report page, cite that URL — not their homepage or about page.
+
+3. **If you can't find a URL for a claim, don't include the claim.** No unsourced facts. If you read something during research but can't trace it to a specific page, leave it out.
+
+4. **Verify numbers match the source exactly.** After writing a claim with a number (percentage, dollar amount, count), re-read the cited page and confirm the exact figure appears there. Common errors: writing 59% when the source says 75%, writing 4,000 when the source says 1,651, or writing 20% when the source says 25%. If your number doesn't match, use the source's number or remove the claim.
+
+5. **Attribution matters.** Say "X reports that" when citing an org's own claims. Say "independent evaluation found" when citing third-party evidence. The distinction is load-bearing.
+
+6. **Format:** `[Source Name](URL)` inline. The SOURCES section at the end must list every URL cited in the report.
+
+### 4. Before-Submission Quality Checks
+
+Run these checks before submitting. They are not optional.
+
+**Structure:**
+- [ ] All 5 prompt tables present and complete (20 rows each)
+- [ ] All 7 summary sections present with substantive content
+- [ ] SOURCES section lists every URL cited inline
+- [ ] Scored checklist adds up correctly
+
+**Citations:**
+- [ ] Every factual claim has its own inline citation
+- [ ] Spot-check at least 5 citations: visit the URL and confirm the EXACT numbers on the page match what you wrote. If the source says 132% and you wrote 136%, fix it.
+- [ ] For any citation where the page doesn't support your claim, find the correct source or remove the claim
+- [ ] No claims are cited to general overview pages when a specific report or data page exists
+
+**Writing style:**
+- [ ] No em dashes (—). Replace with periods, commas, or parentheses.
+- [ ] No filler adjectives: seamless, robust, comprehensive, innovative, cutting-edge, holistic, game-changing
+- [ ] No AI transitions: "It's worth noting", "Here's the thing", "Let's dive in", "Simply put"
+- [ ] Replace "leverage" with "use", "utilize" with "use"
+- [ ] Paragraphs under 4 sentences
+- [ ] No superlatives unless backed by comparative data
+
+### 5. Submit
+
+Submit using `submit_report` with the full markdown as `report_markdown`. Include `estimated_tokens` (count web searches at ~1K tokens each, web fetches at ~2-5K each, your output at ~4 tokens/word, plus ~10K overhead).

package/pipeline/02-verify/PROMPT.md

@@ -1,69 +1,69 @@
-# Verify Citations (Standalone Re-verification)
-
-Use this methodology when re-verifying an existing report. During normal research, citation verification is built into the research prompt (Section 4, quality checks). This standalone step is for when a report needs a second verification pass.
-
-## Instructions
-
-### 1. Read the Report
-
-Read the full research report. Note every inline citation `[Source Name](URL)` and every factual claim.
-
-### 2. Test Every Citation
-
-For each citation, visit the URL using web fetch and verify:
-
-- **URL loads** — Is it a real page (not 404, not a redirect to a homepage)?
-- **Content matches** — Does the page actually say what the report claims? Quote the relevant passage.
-- **Data is accurate** — Do the numbers match?
-
-Record each check:
-
-| # | Citation | URL Status | Content Match | Notes |
-|---|----------|-----------|---------------|-------|
-
-Status values:
-- **VALID** — URL loads and content matches
-- **BROKEN** — 404 or page doesn't load
-- **MISMATCH** — URL loads but doesn't support the claim
-- **PARTIAL** — Some claims match, some don't
-- **UNVERIFIABLE** — Paywalled or content not accessible
-
-### 3. Re-attribute Mismatches
-
-For each MISMATCH or PARTIAL citation:
-1. Use web search to find the correct source for the claim
-2. If found: replace the citation URL with the correct one
-3. If not found anywhere: remove the claim from the report or add a caveat ("This claim could not be independently verified")
-
-Do not leave misattributed citations in place.
-
-### 4. Check for Hallucinations
-
-Search the web for claims that seem unusually specific:
-- Statistics that don't appear in any source
-- Named studies or RCTs that can't be found
-- Program details that contradict other sources
-
-### 5. Apply Corrections
-
-For each issue:
-
-```
-### Correction [N]
-**Location:** [First ~10 words of the problematic passage]
-**Problem:** [What's wrong]
-**Fix:** [What was changed]
-```
-
-### 6. Output
-
-Write the corrected report with a verification summary at the top:
-
-```markdown
-## Verification Summary
-- Citations checked: X
-- Valid: X | Broken: X | Mismatch: X | Partial: X
-- Claims removed (unsourced): X
-- Citations re-attributed: X
-- Corrections applied: X
-```
+# Verify Citations (Standalone Re-verification)
+
+Use this methodology when re-verifying an existing report. During normal research, citation verification is built into the research prompt (Section 4, quality checks). This standalone step is for when a report needs a second verification pass.
+
+## Instructions
+
+### 1. Read the Report
+
+Read the full research report. Note every inline citation `[Source Name](URL)` and every factual claim.
+
+### 2. Test Every Citation
+
+For each citation, visit the URL using web fetch and verify:
+
+- **URL loads** — Is it a real page (not 404, not a redirect to a homepage)?
+- **Content matches** — Does the page actually say what the report claims? Quote the relevant passage.
+- **Data is accurate** — Do the numbers match?
+
+Record each check:
+
+| # | Citation | URL Status | Content Match | Notes |
+|---|----------|-----------|---------------|-------|
+
+Status values:
+- **VALID** — URL loads and content matches
+- **BROKEN** — 404 or page doesn't load
+- **MISMATCH** — URL loads but doesn't support the claim
+- **PARTIAL** — Some claims match, some don't
+- **UNVERIFIABLE** — Paywalled or content not accessible
+
+### 3. Re-attribute Mismatches
+
+For each MISMATCH or PARTIAL citation:
+1. Use web search to find the correct source for the claim
+2. If found: replace the citation URL with the correct one
+3. If not found anywhere: remove the claim from the report or add a caveat ("This claim could not be independently verified")
+
+Do not leave misattributed citations in place.
+
+### 4. Check for Hallucinations
+
+Search the web for claims that seem unusually specific:
+- Statistics that don't appear in any source
+- Named studies or RCTs that can't be found
+- Program details that contradict other sources
+
+### 5. Apply Corrections
+
+For each issue:
+
+```
+### Correction [N]
+**Location:** [First ~10 words of the problematic passage]
+**Problem:** [What's wrong]
+**Fix:** [What was changed]
+```
+
+### 6. Output
+
+Write the corrected report with a verification summary at the top:
+
+```markdown
+## Verification Summary
+- Citations checked: X
+- Valid: X | Broken: X | Mismatch: X | Partial: X
+- Claims removed (unsourced): X
+- Citations re-attributed: X
+- Corrections applied: X
+```