| name | fact-check |
| description | Fact-check content across the ImpactMojo platform — blog posts, courses, labs, games, and book summaries. Use when the user says "fact check", "check facts", "verify content", "review for errors", or on a regular content audit schedule. |
Fact-Check Content Audit
Systematically review ImpactMojo content for factual accuracy. Run this regularly (monthly recommended) or before major launches.
Scope
Review content in this priority order:
- Blog posts (
blog/*.html) — highest risk, public-facing thought leadership
- Flagship courses (
courses/*/index.html, courses/*/lexicon.html) — educational content must be accurate
- 101 courses (
101-courses/*.html) — foundational course decks
- Book summaries (
BookSummaries/*-companion.html) — attributed claims and citations
- Labs (
Labs/*-lab.html) — methodology descriptions and worked examples
- Games (
Games/*.html) — scenario text and factual framing
Steps
-
Gather content files
echo "=== Content inventory ==="
echo "Blog posts: $(ls blog/*.html 2>/dev/null | wc -l)"
echo "Flagship courses: $(ls courses/*/index.html 2>/dev/null | wc -l)"
echo "Lexicons: $(ls courses/*/lexicon.html courses/*/media-lexicon.html 2>/dev/null | wc -l)"
echo "101 courses: $(ls 101-courses/*.html 2>/dev/null | wc -l)"
echo "Book summaries: $(ls BookSummaries/*-companion.html 2>/dev/null | wc -l)"
echo "Labs: $(ls Labs/*-lab.html 2>/dev/null | wc -l)"
echo "Games: $(ls Games/*.html 2>/dev/null | wc -l)"
-
Check for known recurring errors
Run these automated checks first to catch common issues:
grep -rn "Research Rundown.*weekly\|weekly.*Research Rundown" blog/ courses/ --include="*.html"
grep -rn "developed by USAID\|created by USAID\|USAID developed\|USAID created" blog/ courses/ --include="*.html" | grep -i logframe
grep -rn "coined by.*Edmondson\|Edmondson.*coined" blog/ courses/ --include="*.html"
grep -rn "Benjamin Franklin" blog/ courses/ --include="*.html"
grep -rn "OpenCourseWare.*2001.*thousands\|2001.*OpenCourseWare.*thousands" blog/ courses/ --include="*.html"
-
Review content by category
For each content type, use parallel subagents to read article content (skip CSS/JS boilerplate) and flag:
Hard errors (must fix):
- Wrong dates (treaty signed, policy enacted, organisation founded)
- Wrong attributions (quote or concept credited to wrong person)
- Wrong statistics (population figures, percentages, counts)
- Wrong organisation/policy names
- Wrong methodology descriptions (formulas, definitions)
- Outdated facts presented as current (e.g., stale population counts)
Soft issues (flag for review):
- Unverifiable quotes attributed to specific people
- Approximate statistics without source caveats
- Simplifications that could mislead (e.g., "X invented Y" when X popularised Y)
- Outdated but technically-not-wrong figures ("over 400 million" when real number is 600 million)
-
Cross-check internal consistency
grep -rn "OECD DAC.*199[0-9]\|DAC criteria.*199[0-9]" blog/ courses/ --include="*.html"
grep -rn "Sphere.*20[0-9][0-9]\|Sphere Standards.*first" blog/ courses/ --include="*.html"
grep -rn "Paris Declaration.*20[0-9][0-9]" blog/ courses/ --include="*.html"
-
Fix errors
- Fix all hard errors immediately
- For soft issues, add appropriate hedging language ("widely attributed to", "approximately", "as of [year]")
- Document what was changed and why
-
Cross-link audit
While reviewing, also check:
- Blog posts should have 4-7 in-text cross-links to sibling posts and ImpactMojo labs/tools
- Course pages should link to relevant blog posts, labs, and games
- First mention only, relative paths, no self-links
-
Update known-errors registry
After fixing, add new patterns to the automated checks in Step 2 above so they're caught in future runs. Edit this skill file to add new grep patterns.
-
Commit and report
git add -A
git commit -m "Fact-check audit: fix [N] errors across [content types]"
git push
Report to user:
- Number of hard errors found and fixed
- Number of soft issues flagged
- Files modified
- Any items needing human judgment
Known Correct Facts (Reference)
Keep this section updated as a quick-reference for frequently checked claims:
| Claim | Correct Fact | Source |
|---|
| Logframe origin | Defense sector systems analysis (PPBS); formalised by USAID via Practical Concepts Inc., 1969 | Rosenberg (1969) |
| OECD DAC evaluation criteria | First published 1991; updated 2019 (added coherence) | OECD |
| Sphere Standards | First edition 2000 | Sphere Association |
| Paris Declaration | 2005 | OECD |
| Accra Agenda for Action | 2008 | OECD |
| HAP established | 2003 | HAP International |
| Psychological safety | Schein & Bennis (1965); popularised by Edmondson (1999) | Administrative Science Quarterly |
| "Tell me and I forget..." | Disputed — often misattributed to Franklin/Confucius; cf. Xunzi | Quote Investigator |
| MIT OCW | Announced 2001, pilot launched 2002 (~50 courses), grew to 2,400+ | MIT OCW |
| Research Rundown frequency | Bimonthly (once every two months) | Internal |
| Nepal local governments | 753 local levels per 2015 constitution | Constitution of Nepal |
| India scheduled languages | 22 (Eighth Schedule) | Constitution of India |
| Cohen's d conventions | 0.2 small, 0.5 medium, 0.8 large | Cohen (1988) |
| Belmont Report principles | Respect for persons, beneficence, justice (1979) | HHS |
| UNESCO OER Recommendation | Adopted November 2019 | UNESCO |