@human-avatar/skills-for-humanity 1.0.0

package/skills/emotional-resistance-diagnosis/SKILL.md

@@ -0,0 +1,96 @@
+---
+name: emotional-resistance-diagnosis
+description: "Diagnoses why people are resisting — finding what's underneath the pushback. Resistance is information, not obstruction; it always has a source. TRIGGERS: 'why are they resisting', 'diagnose the pushback', 'they won't get on board', 'people aren't buying in', 'resistance diagnosis'."
+---
+
+# Emotional Resistance Diagnosis
+
+Resistance is not a problem to overcome — it is a signal to decode. People resist for
+different reasons, and applying the wrong response to the wrong type makes it worse.
+Presenting more data to someone who is emotionally resistant does nothing. Acknowledging
+feelings with someone who has an intellectual objection is patronising. This skill
+identifies the source before prescribing the response.
+
+---
+
+## Your Process
+
+**Step 1: Describe the Resistance**
+Who is resisting, what are they saying explicitly, and how are they behaving?
+Get behaviorally specific — passive non-compliance, vocal objection, questions
+designed to slow things down, and political manoeuvring are different signals
+pointing to different sources.
+
+**Step 2: Classify Each Instance**
+Assign each type of resistance to one or more of these categories:
+- **Intellectual** — they disagree with the reasoning, evidence, or conclusion.
+  They think you're wrong.
+- **Emotional** — they feel something important to them is at risk. They may not
+  be able to articulate what, but something feels threatening.
+- **Political** — a competing interest is served by the current state. Changing
+  things costs them something real.
+- **Practical** — they don't believe the plan can actually work. They've seen
+  similar things tried and fail.
+
+**Step 3: Source of Each Type**
+Dig to the specific source. Intellectual: which claim do they reject, and why?
+Emotional: what are they afraid of losing — status, security, relationships,
+credit? Political: whose interests benefit from the status quo, and how do they
+intersect with this person? Practical: what specifically do they believe will
+fail, and what informs that belief?
+
+**Step 4: What Are They Legitimately Protecting?**
+Most resistance protects something real — a working system, a relationship, a
+principle, an investment of time and reputation. Name it explicitly. Even if their
+expression of resistance is unhelpful, the thing they're protecting may be
+legitimate. Dismissing this hardens resistance into permanent opposition.
+
+**Step 5: What Would Reduce Each Type?**
+Each type has a different lever. Intellectual resistance needs evidence, logic,
+or acknowledgment that their counter-argument was considered. Emotional needs
+acknowledgment of the risk they feel and credible reassurance. Political needs
+negotiation or incentive realignment. Practical needs proof of concept or a staged
+rollout that limits downside.
+
+**Step 6: Engagement vs Clarity**
+Some resistance requires genuine engagement — changing the plan, negotiating
+trade-offs, addressing real concerns with merit. Some requires clarity — the plan
+isn't actually what they think it is, and better communication resolves it.
+Distinguish these before acting.
+
+---
+
+## Human Check-in
+
+Before proceeding, ask the user:
+
+**How do you want to run this?**
+
+- **A) Full analysis** — complete all steps, reasoning shown throughout
+- **B) Key findings only** — bottom-line output, skip step-by-step detail
+- **C) Resistance source only** — what's actually causing the resistance, skip the implications
+- **D) Refine the framing** — adjust what we're analyzing before starting
+
+Proceed based on their choice.
+
+## Output Format
+
+**Resistance Diagnosis**
+
+| Resistance | Type | Source | What They're Legitimately Protecting | What Reduces It | Engage or Clarify? |
+|---|---|---|---|---|---|
+| [description] | [type] | [root cause] | [legitimate interest] | [lever] | [engage/clarify] |
+
+**Engagement Priority**
+Which resistance requires the most substantive response first, and why. Name the
+one thing that, if addressed well, would most shift the overall dynamic.
+
+---
+
+## Notes
+
+The worst response to resistance is to push harder. The best is to correctly
+classify it first — then respond to the actual type. Treating emotional resistance
+as intellectual (presenting more data) is a common and costly mistake. Treating
+political resistance as emotional (offering reassurance) is equally ineffective.
+Classification is the work.

package/skills/emotional-stakes-mapping/SKILL.md

@@ -0,0 +1,98 @@
+---
+name: emotional-stakes-mapping
+description: "Maps what each stakeholder actually cares about underneath their stated position — because addressing the stated position while missing the real stake accomplishes nothing. TRIGGERS: 'what do they actually want', 'underlying stakes', 'what's really at stake', 'beneath the position', 'why won't they agree'."
+---
+
+# Emotional Stakes Mapping
+
+People argue about positions. They care about stakes. Addressing the stated position
+while missing the real stake is the most common reason negotiations, decisions, and
+alignment efforts fail. This skill maps the gap between the two — not by guessing,
+but by working systematically through what each party stands to lose if the outcome
+goes against them.
+
+---
+
+## Your Process
+
+**Step 1: List Stakeholders**
+Identify every party with a meaningful stake in the outcome — including silent ones
+who will be affected but aren't in the room. Passive non-participants often have the
+highest stakes; they just have no forum to name them.
+
+**Step 2: Stated Position**
+What is each stakeholder explicitly asking for, pushing toward, or resisting? Keep
+this behaviorally specific — "they want sign-off authority" not "they want control."
+Surface the precise ask.
+
+**Step 3: Real Fear**
+What are they actually afraid of losing? Common real fears: status within their
+organisation, control over resources they currently hold, credit for outcomes they've
+invested in, relevance if the change makes their expertise obsolete, relationships
+that depend on the current arrangement, safety from blame if things go wrong. Fears
+drive positions — find the fear beneath the ask.
+
+**Step 4: Minimum Conditions for Agreement**
+What would need to be true — not just what would help, but what is the floor — for
+each stakeholder to feel safe enough to say yes? The minimum condition is often
+surprisingly specific once the real fear is named: "I need to be cited as a
+contributor" or "I need a fallback option if this fails."
+
+**Step 5: Face-Saving Explanation**
+How would they explain agreement to their team, their boss, or themselves? A
+stakeholder may privately accept a compromise but publicly need a narrative that
+doesn't look like capitulation. Agreement that can't be explained is agreement that
+won't hold — or won't be delivered.
+
+**Step 6: Map Alignment and Conflict Zones**
+Find where underlying stakes actually overlap (alignment zones) and where they
+genuinely compete (conflict zones). Alignment at the stake level often exists even
+when surface positions look completely opposed. Conflict zones where one party's win
+structurally requires another's loss cannot be resolved through reframing — they
+require explicit negotiation.
+
+---
+
+## Human Check-in
+
+Before proceeding, ask the user:
+
+**How do you want to run this?**
+
+- **A) Full analysis** — complete all steps, reasoning shown throughout
+- **B) Key findings only** — bottom-line output, skip step-by-step detail
+- **C) Highest stakes only** — what each person cares about most, skip secondary concerns
+- **D) Refine the framing** — adjust what we're analyzing before starting
+
+Proceed based on their choice.
+
+## Output Format
+
+**Stakeholder Table**
+
+| Stakeholder | Stated Position | Real Fear | Minimum Condition | Face-Saving Explanation |
+|---|---|---|---|---|
+| [name/role] | [explicit ask] | [underlying fear] | [the floor for agreement] | [how they'd narrate agreement] |
+
+**Alignment Zones**
+Stakes that multiple stakeholders share underneath apparently conflicting positions.
+List the shared underlying interest and which parties share it.
+
+**Conflict Zones**
+Stakes that genuinely compete — where one party's win requires another's loss. Name
+each conflict zone explicitly and identify what explicit negotiation or trade-off it
+requires.
+
+**Next Move**
+Given the alignment and conflict zones: what is the highest-leverage first
+conversation or action?
+
+---
+
+## Notes
+
+Surface-level positions are negotiating stances; underlying stakes are the actual
+terrain. Work on the terrain. Conflict zones that can't be resolved through alignment
+require explicit trade-off decisions — don't paper over them with language that
+pretends everyone wins when they don't. The face-saving explanation step is often
+what makes agreement stick in practice even after it's been reached in principle.

package/skills/emotional-trust-audit/SKILL.md

@@ -0,0 +1,96 @@
+---
+name: emotional-trust-audit
+description: "Maps what is building and eroding trust in a relationship or situation — trust degrades silently until it fails loudly. TRIGGERS: 'trust audit', 'why don't they trust us', 'relationship health check', 'what's eroding trust', 'build trust with'."
+---
+
+# Emotional Trust Audit
+
+Trust does not fail suddenly — it erodes incrementally through small signals that
+accumulate below the surface. By the time distrust becomes visible in behaviour it
+is usually already deep. The four drivers of trust — competence, reliability,
+integrity, benevolence — degrade independently, and repair requires identifying which
+driver is most damaged rather than applying generic trust-building gestures that
+target the wrong deficit.
+
+---
+
+## Your Process
+
+**Step 1: Name the Relationship**
+Specify the relationship being audited — two individuals, a team and its stakeholders,
+a vendor relationship, a product and its users. Be precise about direction: whose
+trust in whom is being assessed?
+
+**Step 2: Assess Each Driver**
+For each of the four trust drivers, identify recent concrete evidence on both sides.
+Evidence must be specific — named events, observed behaviours, cited decisions.
+Generalisations don't diagnose.
+
+- **Competence** — Can they do what they say? Evidence for: delivered results,
+  demonstrated expertise, track record. Evidence against: failures, skill gaps,
+  over-promising relative to delivery.
+- **Reliability** — Do they do what they say? Evidence for: consistent follow-through,
+  keeping commitments under pressure. Evidence against: broken commitments, dropped
+  items, variable responsiveness.
+- **Integrity** — Do they act in line with their stated values? Evidence for:
+  transparent communication when it's costly, decisions consistent across contexts.
+  Evidence against: values invoked selectively, principles abandoned under pressure.
+- **Benevolence** — Do they have my interests at heart? Evidence for: advocacy on
+  behalf of the other party, proactive disclosure of relevant information. Evidence
+  against: decisions made without considering impact, information withheld.
+
+**Step 3: Identify the Most Damaged Driver**
+Where is the deficit largest relative to what the relationship requires to function?
+This is the repair priority — not where there's the most noise, but where the
+structural deficit is deepest.
+
+**Step 4: Highest-Leverage Trust-Building Action**
+What single specific act would most credibly signal strength on the most damaged
+driver? The action must be proportionate and credible — visible enough to register,
+consistent enough to be believed, not so large it looks performative.
+
+**Step 5: Behaviour to Stop**
+What is actively eroding trust right now that must stop before any repair action
+can land? Repair actions launched while trust-eroding behaviour continues produce
+no signal — the erosion cancels the repair.
+
+---
+
+## Human Check-in
+
+Before proceeding, ask the user:
+
+**How do you want to run this?**
+
+- **A) Full analysis** — complete all steps, reasoning shown throughout
+- **B) Key findings only** — bottom-line output, skip step-by-step detail
+- **C) Trust deficits only** — what's actively eroding trust right now
+- **D) Refine the framing** — adjust what we're analyzing before starting
+
+Proceed based on their choice.
+
+## Output Format
+
+**Driver Assessment**
+
+| Driver | Evidence Building Trust | Evidence Eroding Trust | Net Assessment |
+|---|---|---|---|
+| Competence | [specific evidence] | [specific evidence] | [strong/neutral/deficit] |
+| Reliability | [specific evidence] | [specific evidence] | [strong/neutral/deficit] |
+| Integrity | [specific evidence] | [specific evidence] | [strong/neutral/deficit] |
+| Benevolence | [specific evidence] | [specific evidence] | [strong/neutral/deficit] |
+
+**Most Damaged Driver:** [driver + one-sentence rationale]
+
+**Highest-Leverage Action:** [specific, concrete act — what, by whom, by when]
+
+**Behaviour to Stop:** [what must cease for repair actions to have any effect]
+
+---
+
+## Notes
+
+Trust is rebuilt through consistent small acts, not dramatic gestures. The action
+recommended here is a first step, not a solution — sustainable repair requires
+sustained consistency. The behaviour-to-stop is non-negotiable and must precede
+everything else; repair while erosion continues is wasted effort.

package/skills/ethics/SKILL.md

@@ -0,0 +1,130 @@
+---
+name: ethics
+description: "Entry point for the ethics toolkit. Routes to the right ethical analysis skill based on your situation. Use when you say 'ethics', 'is this ethical', 'sanity check this', 'who does this affect', 'is this fair', 'something feels wrong', or want ethical reasoning applied without knowing which specific tool fits."
+---
+
+# Ethics
+
+Applies ethical reasoning to decisions, designs, policies, and practices. Diagnoses the type and depth of ethical work needed and applies the right tool.
+
+## Which tool fits
+
+| You need to... | Tool |
+|---|---|
+| Comprehensive multi-framework council with peer review | ethics-council |
+| Fast complete ethics report across all frameworks | ethics-check |
+| Rapid ethical response to an active incident | ethics-crisis-triage |
+| Audit a data collection or sharing decision | ethics-data-audit |
+| Check an algorithm or model for bias and fairness | ethics-bias-check |
+| Review a UX flow for genuine consent | ethics-consent-review |
+| Quick ethical impact scan before shipping | ethics-impact-scan |
+| Evaluate a vendor or supplier against ethical standards | ethics-vendor-review |
+
+## Routing Decision
+
+- **High-stakes decision affecting many stakeholders, wants thorough pressure-testing** → ethics-council (full council with peer review, HTML report)
+- **Needs a complete ethical assessment quickly** → ethics-check (all 5 frameworks, no peer review overhead)
+- **Something has already gone wrong — active incident** → ethics-crisis-triage
+- **Data collection, retention, or sharing decision** → ethics-data-audit
+- **Algorithm, ML model, ranking, or scoring system** → ethics-bias-check
+- **Checkout flow, onboarding, consent, dark patterns** → ethics-consent-review
+- **About to ship something, quick impact check** → ethics-impact-scan
+- **Evaluating a third-party vendor, API, or partner** → ethics-vendor-review
+- **Unclear** → ethics-check (comprehensive but lightweight — surfaces which deeper tool is needed)
+
+## Confirm Direction
+
+After diagnosing which tool fits, present the recommendation before executing:
+
+> My read: **[diagnosed tool]** — one sentence on why it fits.
+
+- **A) Yes, run that tool**
+- **B) Show me all options** — list every skill in this category with one-line descriptions
+- **C) Quick version** — lighter-weight alternative for this situation, if one exists
+- **D) Re-diagnose** — describe the situation differently for a second read
+
+Wait for their selection before proceeding.
+
+---
+
+## Ethics Check
+
+*Fast comprehensive ethics report across all five frameworks.*
+
+Run the situation through all five ethical frameworks in a single pass:
+
+1. **Utilitarian:** Who is affected and how? Does this maximize net benefit across all parties?
+2. **Deontological:** Are any duties being violated or rights being overridden, regardless of outcomes?
+3. **Virtue Ethics:** What does this decision say about character? Would someone of integrity do this?
+4. **Care Ethics:** Who is vulnerable or in a dependent relationship? Are we honoring those dependencies?
+5. **Justice/Fairness:** Is this fair to everyone, including those with the least power?
+
+Synthesize: where do the frameworks agree (high-confidence signal)? Where do they conflict (genuine value tension that must be owned)? Issue a verdict with a direct recommendation.
+
+---
+
+## Ethics Council
+
+*Full five-advisor council with peer review — use for high-stakes decisions.*
+
+See `ethics-council` for the full multi-agent process with 5 independent framework advisors, peer review, chair synthesis, and HTML report generation. Route here when the stakes are high enough to warrant that depth.
+
+---
+
+## Ethics Crisis Triage
+
+*Rapid ethical assessment when something has already gone wrong.*
+
+In a crisis, the instinct is to manage rather than reason. This tool forces ethical clarity under pressure. Apply a rapid three-layer assessment: (1) Immediate harm — who is being harmed right now, and what stops it fastest? (2) Accountability — what transparency is owed, and to whom? (3) Response ethics — which response options themselves create new ethical problems? Crisis responses that cut ethical corners tend to create second crises.
+
+**Output:** Immediate harm assessment, accountability obligations, ethically viable response options, and what to avoid.
+
+---
+
+## Ethics Data Audit
+
+*Audit a data decision against ethical standards.*
+
+Goes beyond legal compliance. Assess: (1) Necessity — is collecting this data actually required for the stated purpose? (2) Proportionality — is the scope of collection proportional to the benefit? (3) Consent — do people meaningfully understand and agree to this? (4) Harm potential — what's the worst realistic use or breach? (5) Retention — how long is too long?
+
+**Output:** Data practice assessment across all five dimensions, with specific changes that would make the practice clearly ethical.
+
+---
+
+## Ethics Bias Check
+
+*Evaluates an algorithm or model for discriminatory patterns.*
+
+Assess the system against protected characteristics: does it produce systematically different outcomes for different groups? If so: is the difference justified by legitimate criteria, or does it reflect historical bias in training data? Apply four fairness standards: equal treatment, equal outcome, individual fairness, and counterfactual fairness. Flag which standard is being violated and by what mechanism.
+
+**Output:** Fairness assessment per standard, identified disparate impacts, root cause of bias, and mitigation recommendations.
+
+---
+
+## Ethics Consent Review
+
+*Reviews a UX flow for genuine consent.*
+
+Genuine consent is informed, voluntary, and meaningful — not just legally checkboxed. Evaluate the flow: Can users meaningfully understand what they're agreeing to? Are they under pressure (dark patterns, urgency, buried choices)? Is opting out as easy as opting in? Would a user feel deceived after the fact?
+
+**Output:** Consent quality assessment, specific dark patterns identified, and changes needed for consent to be genuine.
+
+---
+
+## Ethics Impact Scan
+
+*Quick pre-ship ethical impact assessment.*
+
+Before committing, surface: who benefits from this? Who bears costs or risk? At what scale? Are the costs and benefits distributed fairly? Are any harms irreversible? This is a lightweight check that takes minutes — the goal is to catch obvious ethical problems before they become embedded in shipped product.
+
+**Output:** Benefit/harm map, distribution assessment, irreversibility flags, and a go/no-go recommendation with conditions.
+
+---
+
+## Ethics Vendor Review
+
+*Evaluates a third-party against ethical standards.*
+
+Assess across five domains: (1) Labor practices — how are workers treated in the supply chain? (2) Data handling — how do they use data about you and your users? (3) Business model — does their incentive structure align with your values? (4) Environmental impact, (5) Political/social alignment — do their positions or practices conflict with stated values? A vendor whose practices conflict with your values makes your product complicit.
+
+**Output:** Vendor assessment across all five domains, red flags, and a recommendation on whether to proceed.

package/skills/ethics-bias-check/SKILL.md

@@ -0,0 +1,90 @@
+---
+name: ethics-bias-check
+description: "Evaluate an algorithm, model, ranking system, recommendation engine, or automated decision process for discriminatory patterns and unfair outcomes. Use before deploying any system that makes decisions about people or ranks/filters/scores people. TRIGGERS: 'bias check', 'check this algorithm for bias', 'is this fair', 'fairness audit', any ML model, ranking function, scoring system, content recommendation, or automated decision that differentiates between people."
+---
+
+# Ethics Bias Check
+
+Algorithms that treat everyone the same can still discriminate. A ranking that optimises for engagement may systematically deprioritise certain groups. A model trained on historical data may encode historical injustice. A feature that works well on average may fail badly for users who aren't the implicit default.
+
+This check surfaces those patterns before they ship.
+
+---
+
+## Your Process
+
+**Step 1: Define the system**
+What is the algorithm, model, or automated decision? What is its input? What is its output? Who does it make decisions *about*? What happens to people based on its output?
+
+**Step 2: Identify the implicit default**
+Every system has a default user in mind — often implicitly. Ask:
+- Who was this optimised for?
+- Whose behaviour or data was used to train or calibrate it?
+- Who is absent from the training set or design process?
+
+The implicit default is often the demographic that experiences least friction. Others bear the cost of that assumption.
+
+**Step 3: Check for direct bias**
+Does the system use protected characteristics (age, gender, race, disability, location as proxy for race, etc.) as features, or correlates that map closely to them? Does it produce different outcomes for different demographic groups? Is that difference *justified* (e.g. a medical dosage model that accounts for body weight) or *unjustified* (e.g. a loan model that penalises postcodes that correlate with race)?
+
+**Step 4: Check for proxy bias**
+Protected characteristics don't need to be explicit to create discriminatory outcomes. Audit the features used:
+- Which features correlate with protected characteristics?
+- Could the model achieve similar accuracy without the proxy, and if so, is there a reason it uses it?
+- What's the *historical source* of the training data? Does it encode past discriminatory practices?
+
+**Step 5: Check the feedback loop**
+Many deployed systems create the conditions that confirm their own predictions. A recommendation system that shows users content they're predicted to engage with can entrench filter bubbles. A fraud model that flags more accounts from a certain demographic leads to more investigations of that demographic, which produces more evidence that "justifies" the bias. Ask: does this system create or amplify the patterns it's trying to predict?
+
+**Step 6: Assess the harm distribution**
+When the system makes errors, who bears the cost?
+- False positives and false negatives are not equally distributed — who experiences more of each?
+- What happens to a person when the system is wrong about them? Is the cost significant?
+- Do people have recourse? Can they appeal or correct the system's decision about them?
+
+---
+
+## Human Check-in
+
+Before proceeding, ask the user:
+
+**How do you want to run this?**
+
+- **A) Full analysis** — complete all steps, reasoning shown throughout
+- **B) Key findings only** — bottom-line output, skip step-by-step detail
+- **C) Disparate impact only** — flag differential outcomes without full root-cause analysis
+- **D) Refine the framing** — adjust what we're analyzing before starting
+
+Proceed based on their choice.
+
+## Output Format
+
+**System Being Audited:**
+[What it does, inputs, outputs, who it affects]
+
+**Implicit Default**
+[Who the system is optimised for; who may be disadvantaged by that assumption]
+
+**Bias Findings**
+
+| Dimension | Finding | Severity |
+|---|---|---|
+| Direct bias (protected characteristics) | [finding] | 🔴 / 🟡 / 🟢 |
+| Proxy bias (correlated features) | [finding] | 🔴 / 🟡 / 🟢 |
+| Training data issues | [finding] | 🔴 / 🟡 / 🟢 |
+| Feedback loop risk | [finding] | 🔴 / 🟡 / 🟢 |
+| Error cost distribution | [finding] | 🔴 / 🟡 / 🟢 |
+
+**Most Significant Concern**
+[One specific, concrete finding that warrants most attention]
+
+**Recommended Actions**
+- [Specific mitigation per finding, or "escalate to ethics-council" for significant concerns]
+
+---
+
+## Notes
+
+A clean bias check is not a guarantee of fairness — it is evidence of serious effort. Bias is often subtle and emerges at scale. Where this check surfaces concerns, treat them as decisions to be made consciously, not problems to be explained away.
+
+For systems with high-stakes outputs (credit, hiring, healthcare, content moderation), this check is a minimum. Consider ongoing monitoring post-deployment, not just a pre-ship audit.

package/skills/ethics-check/SKILL.md

@@ -0,0 +1,86 @@
+---
+name: ethics-check
+description: "A fast, comprehensive ethics report on any decision, action, or situation — runs all five ethical frameworks in a single pass. Lighter than ethics-council (no peer review, no HTML report), heavier than ethics-impact-scan. Triggers: 'ethics check', 'quick ethics review', 'full ethics report', 'check this ethically', 'run an ethics check', any request for a complete ethical assessment without full council process."
+---
+
+# Ethics Check
+
+Five ethical frameworks consistently applied produce a more complete picture than any single one. Utilitarian analysis may miss rights violations that deontology catches. Virtue ethics surfaces character questions that consequentialism ignores. Running all five in sequence forces a complete assessment and reveals where frameworks agree and where they pull in different directions.
+
+---
+
+## Your Process
+
+**Step 1: State the Decision or Action Clearly**
+Before analysis: name exactly what is being evaluated. A vague subject produces vague ethics. Include who does what, to whom, under what conditions.
+
+**Step 2: Run Each Framework in Sequence**
+
+**Utilitarian** — Net effect on wellbeing across all affected parties. Who benefits, who is harmed, by how much, and how certain are these outcomes? Does the action produce the greatest good for the greatest number? What are the second-order effects?
+
+**Deontological** — Duties and rights. Does this action treat anyone merely as a means to an end? Does it violate any duty — honesty, fairness, non-harm — regardless of outcomes? Are any rights being overridden without adequate justification?
+
+**Virtue Ethics** — Character and integrity. What does this action say about the character of the person or organisation taking it? Is this what a person of genuine integrity and practical wisdom would do? Does it reflect the virtues that matter in this domain?
+
+**Care Ethics** — Relationships and vulnerability. Who is in a relationship of care or dependency here? Does the action honour those relationships? Does it adequately attend to those who are most vulnerable?
+
+**Justice/Fairness** — Distribution and procedure. Is the distribution of benefits and burdens fair? Were affected parties included in the process? Would this decision be defensible behind a veil of ignorance — not knowing which party you would be?
+
+**Step 3: Synthesise**
+Where do the frameworks agree? That agreement is strong ethical signal. Where do they conflict? Name the specific values in tension.
+
+**Step 4: Recommend**
+Given the synthesis: what should happen, and what conditions or safeguards matter most?
+
+---
+
+## Human Check-in
+
+Before proceeding, ask the user:
+
+**How do you want to run this?**
+
+- **A) Full analysis** — complete all steps, reasoning shown throughout
+- **B) Key findings only** — bottom-line output, skip step-by-step detail
+- **C) Framework conflicts only** — where the five frameworks disagree, not where they agree
+- **D) Refine the framing** — adjust what we're analyzing before starting
+
+Proceed based on their choice.
+
+## Output Format
+
+### Decision or Action
+[Clear statement]
+
+### Framework Assessments
+
+**Utilitarian**
+[3–5 sentences: net effects, beneficiaries, harms, certainty, second-order effects]
+
+**Deontological**
+[3–5 sentences: duties, rights, treatment of persons, inviolable constraints]
+
+**Virtue Ethics**
+[3–5 sentences: character revealed, practical wisdom, integrity]
+
+**Care Ethics**
+[3–5 sentences: relationships at stake, vulnerability, responsiveness to those who depend on this]
+
+**Justice / Fairness**
+[3–5 sentences: distribution of benefits and burdens, procedural fairness, veil of ignorance test]
+
+### Agreement and Conflict Summary
+- **Frameworks agree:** [where multiple frameworks converge]
+- **Frameworks conflict:** [where they pull in different directions — name the tension]
+
+### Values at Stake
+- [The core values in tension or at risk]
+
+### Recommendation
+[Clear recommendation with rationale — and any conditions or safeguards that change the assessment]
+
+---
+
+## Notes
+
+Use ethics-council when the situation requires deeper deliberation, peer challenge between frameworks, or a formal report. Use ethics-impact-scan for a lighter first pass. This skill sits between — a complete, fast, single-pass assessment.