Purpose
Take any education program, policy, or funding decision and apply a disciplined equity analysis that names who actually benefits, who gets left out, what harms could surface for marginalized groups, and what concrete changes would re-center the most excluded. Replaces vague "we care about equity" intentions with a documented, defensible evaluation a program designer or funder can act on.
When to Use
Invoke this skill when the user asks to:
- Evaluate an education program or intervention for equity impact
- Screen a grant, funding decision, or RFP against equity criteria
- Stress-test a policy or reform against the populations it claims to serve
- Identify who is being excluded by an existing program's design or delivery
- Assess unintended harms before launching or scaling an initiative
- Redesign a program to better reach girls, low-income, rural, caste-marginalized, disabled, or non-dominant-language learners
Inputs Required
Before starting, collect:
- The intervention: What the program, policy, or funding decision is — purpose, mechanism, and scale
- Stated goal: The outcome it intends to produce and for whom
- Target population: Who it is officially designed to reach
- Delivery context: Geography, education level, urban/rural, and the system it operates within (public school, nonprofit, government scheme, etc.)
- Resources & access requirements: What a participant needs to benefit (cost, device, transport, literacy, time, documents, parental consent)
- Equity dimensions of concern: Which axes matter most here — gender, income, caste, race/ethnicity, disability, language, migration status, geography
- Decision at stake: Fund / don't fund, launch / revise, scale / hold
If equity dimensions of concern are not specified, evaluate across all of: gender, income, caste/ethnicity/race, disability, language, and geography — and flag the two or three most consequential for this context.
Instructions
Step 1: Restate the Intervention and Its Equity Claim
Summarize the intervention in 2–3 sentences in plain language. Then state explicitly what equity claim it is making — implicit or explicit — e.g., "This claims to expand access to STEM for rural girls." Name the claim precisely so the rest of the analysis can test it. If no equity claim exists, note that the intervention is equity-neutral by design and proceed to surface its distributional effects anyway.
Step 2: Map Who Benefits (Beneficiary Map)
Identify who actually gains, ranked by how much benefit they capture:
- Primary beneficiaries: Who the design most effectively reaches
- Secondary beneficiaries: Who benefits indirectly or partially
- Already-advantaged capture: Whether better-resourced groups are positioned to absorb a disproportionate share (e.g., families who already have devices, transport, or information access) For each group, note why the design favors them — the specific access requirement or mechanism that makes benefit easy for them.
Step 3: Identify Who Is Excluded (Exclusion Analysis)
Walk each access requirement from the inputs against each equity dimension. For every requirement, ask: who cannot meet it, and why?
- Cost barriers: Fees, materials, indirect costs (transport, lost income)
- Information barriers: Who hears about it and who doesn't (language of outreach, channels used)
- Eligibility barriers: Documents, residency, prior enrollment, age gates that exclude marginalized groups
- Logistical barriers: Location, timing, distance, child/elder care, mobility
- Social barriers: Norms restricting girls' movement, caste-based exclusion, stigma around disability
- Design barriers: Language of instruction, assumed literacy, assumed connectivity For each excluded group, state the specific barrier and the equity dimension(s) it maps to. Be concrete — name the population, not "the underserved."
Step 4: Surface Unintended Harms (Harm Register)
Identify plausible negative effects on marginalized populations even when intentions are good:
- Widening gaps: Could the intervention raise outcomes for the advantaged faster, increasing relative inequality?
- Resource diversion: Does it pull funding, staff, or attention from existing supports the most marginalized depend on?
- Stigma or targeting harm: Does identifying a group for support expose them to shame, surveillance, or backlash?
- Displacement: Does it crowd out community-led or informal provision that was reaching the excluded?
- Conditionality harm: Do requirements (attendance, performance, behavior) penalize those facing structural barriers? For each, rate Likelihood (Low/Medium/High) and Severity (Low/Medium/High), and name the affected group.
Step 5: Equity Scorecard
Score the intervention on a 1–5 scale across five criteria, with a one-line justification for each:
- Reach: Does it actually arrive at the most marginalized, or stop at the moderately disadvantaged?
- Targeting precision: Does benefit concentrate where need is greatest, or leak to the already-advantaged?
- Barrier removal: Does it dismantle structural barriers, or assume them away?
- Voice & agency: Are affected communities involved in design and feedback, or only recipients?
- Do-no-harm: Net assessment from the harm register Provide a total (out of 25) and a one-sentence verdict (e.g., "Equity-promising but high exclusion risk for rural girls without redesign").
Step 6: Prioritized Redesign Recommendations
Produce 5–8 specific changes that would center the most excluded populations, ranked by expected equity gain relative to effort. For each:
- The change: What to do, concretely
- Who it includes: The specific population it brings in
- The barrier it removes: Tied back to Step 3
- Trade-off: Cost, complexity, or political friction it introduces Lead with changes that remove the highest-impact exclusion barriers identified in Step 3.
Step 7: Decision Recommendation
Make an explicit call on the decision at stake (fund/revise/launch/hold) with a clear rationale grounded in the scorecard and harm register. State the conditions under which the recommendation would change. Be opinionated — name the single most important redesign condition before the intervention should proceed.
Output Format
- Structure: Seven sections following the steps; beneficiary map and exclusion analysis as labeled lists; harm register and scorecard as tables; recommendations as a ranked list
- Length: 1,100–1,500 words
- Format: Structured evaluation document — scannable, with tables for the harm register and equity scorecard
- Sections: Intervention & Equity Claim → Beneficiary Map → Exclusion Analysis → Harm Register → Equity Scorecard → Redesign Recommendations → Decision Recommendation
- Tone: Equity evaluator — rigorous, specific, and willing to name uncomfortable distributional truths; never moralizing, always actionable
Quality Checklist
Before delivering output, verify:
- [ ] Every excluded group is named specifically with the barrier and equity dimension it maps to — no "the underserved" without specifics
- [ ] The harm register rates both likelihood and severity for each harm and names the affected group
- [ ] The equity scorecard scores all five criteria with a one-line justification and a numeric total out of 25
- [ ] Redesign recommendations are ranked by equity gain and each names the specific population it includes
- [ ] Already-advantaged capture is explicitly assessed, not assumed absent
- [ ] A clear decision recommendation is made with stated conditions — no hedging
Notes
- The most common equity failure is benefit leakage to the moderately disadvantaged while the most marginalized remain unreached — interrogate Step 2's "already-advantaged capture" hardest.
- Equity-neutral designs are rarely neutral in effect; surface distributional impact even when no equity claim is made.
- Intersectionality matters: a low-income disabled girl in a non-dominant language community faces compounded barriers no single dimension captures — flag stacked exclusions explicitly.
- If community voice is absent from the design, that alone is grounds to lower the Voice & Agency score regardless of stated good intentions.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills