@aria_asi/cli 0.2.40 → 0.2.41

package/skills/aria-cognition/aria-business-audit/SKILL.md

@@ -0,0 +1,133 @@
+---
+name: aria-business-audit
+description: TRIGGER when the owner asks for a business audit, health check, KPI review, growth diagnostic, financial diagnostic, GTM diagnostic, retention diagnostic, ops audit, team audit, or any "what's broken / what's compounding / what should we fix first" diagnostic across the business. Produces a ranked diagnostic report with the dominant constraint, the highest-leverage fix, and a 30/60/90 plan. Composes mizan + ladunni-22 + ghazali-8lens + predictor + noor-recognition with an audit-specific bottleneck-leverage-evidence frame.
+---
+
+# Aria Business Audit
+
+The cognition skill for end-to-end business diagnostics. Most owners feel multiple things broken simultaneously and don't know which is upstream. This skill ranks the constraints by leverage and produces a concrete fix sequence.
+
+## Prime Doctrine
+
+The right business diagnostic ranks ONE dominant constraint, not seven things to improve.
+
+- **Constraints, not laundry lists.** Theory-of-constraints applies — the throughput of a business is set by the dominant bottleneck. Fixing anything else is local optimization that doesn't compound.
+- **Leverage over breadth.** A 1× fix on the dominant constraint beats 5× fixes on non-constraints. Identify the constraint first, then size the fix.
+- **Evidence beats narrative.** Owner narratives ("our churn is bad") often misidentify the constraint. Cohort data + customer interviews + funnel measurement is the only reliable diagnostic source.
+- **Fixes have shapes.** Some constraints require a campaign (24-72h); others require a system (4-12 weeks); others require a strategic pivot (quarters). Size the fix to the shape, don't promise quarters of work in a 24h motion.
+
+## Trigger Detection
+
+Fire when:
+
+- Owner asks "what's wrong with the business" / "where should we focus" / "audit this"
+- Owner says a metric is broken (churn, conversion, growth, retention, CAC, runway)
+- A KPI review is on the calendar (board prep, investor update, retro)
+- A new fiscal period is starting (planning, OKRs, target-setting)
+- An owner is making a major capital allocation (hire, vendor, channel investment)
+
+## Required Workflow (bottleneck → leverage → evidence → fix-shape)
+
+Every business audit produces:
+
+1. **The funnel skeleton.** Acquisition → activation → retention → expansion → referral. Map the customer's actual path with current numbers per stage.
+2. **The constraint.** Which stage has the lowest throughput relative to a comparable benchmark? The constraint is the stage whose lift would compound most.
+3. **The evidence for the constraint.** Cohort data, interview signal, comparison to comparable companies. NOT "my gut says retention is the problem."
+4. **The leverage size.** If the constraint is fixed to benchmark, what does revenue / runway / NRR look like? Numeric, not directional.
+5. **The fix shape.** Campaign (24-72h, low cost, reversible), system (4-12 weeks, medium cost, harder to reverse), pivot (quarters, high cost, mostly irreversible). Match the shape to the constraint.
+6. **The 30/60/90 plan.** What lands in each window if the fix is the right one. Includes a kill-criterion: at what 30-day signal do we abandon and re-diagnose?
+
+## Audit-Mapped Lenses (`ghazali-8lens`)
+
+- **truth** — is the data current, segmented, and complete? Are we measuring the right metric or a vanity metric?
+- **harm** — does the proposed fix risk customer / team / brand trust if it succeeds at scale?
+- **trust** — would the team / customers / investors recognize this diagnosis as honest?
+- **power** — does the fix shift leverage to us (durable advantage) or rent it (short-term lift, ongoing cost)?
+- **reflection** — would the owner recognize this as their company's real problem, not a consultant's checklist?
+- **context** — is the constraint structural (durable) or temporary (market / season)?
+- **impact** — predicted KPI delta with confidence band, by 30/60/90
+- **beauty** — does the fix feel inevitable to the owner, or a list of nice-to-haves?
+
+## User-Facing Layout (per `aria-readable-output`)
+
+```
+## [Diagnosis in one line — e.g. "Activation is the constraint; day-7 first-action drop is killing CAC payback"]
+
+### Funnel skeleton (current)
+- Acquisition: [number]
+- Activation: [number] ← constraint
+- Retention: [number]
+- Expansion: [number]
+- Referral: [number]
+
+### The constraint, with evidence
+- [Which stage + why + cohort data + benchmark comparison]
+
+### Leverage if fixed
+- [Numeric: ARR / NRR / payback delta if constraint moves to benchmark]
+
+### Fix shape + 30/60/90
+- [Campaign / system / pivot]
+- 30 days: [concrete deliverables + kill criterion]
+- 60 days: [concrete deliverables]
+- 90 days: [concrete deliverables]
+
+**Next:** [first 7-day action with measurable signal]
+```
+
+`<gate>` block when the audit triggers a public commitment, a hire, or a contract change.
+
+## Composition Rule
+
+- Composes with `aria-business-frame` as the substrate
+- Calls `ladunni-22` for the cross-domain pass — most diagnostics surface as one metric but live in multiple domains
+- Calls `mizan` for intent / evidence / protection / quality
+- Calls `tadabbur` for second-order consequences of each candidate fix
+- Calls `ghazali-8lens` per defaults
+- Calls `predictor` for the 30/60/90 predicates
+- Calls `noor-recognition` for prior diagnostic patterns and Hamza's prior decisions
+- Hands off to `aria-revenue-engine` / `aria-gtm-architect` / `aria-retention-engine` once the constraint is named (the relevant vertical owns the fix)
+- Calls `aria-research-orchestrator` if external benchmarks are needed and not in memory
+
+## Recovery Contract
+
+- If funnel data is missing: name the gap, recommend the smallest queries / exports needed, and produce a candidate diagnosis with explicit confidence weight. Do NOT fabricate numbers.
+- If the constraint is genuinely ambiguous (two stages tied): recommend a 14-day measurement sprint to disambiguate before the fix.
+- If the owner narrative contradicts the data: surface both, recommend an interview sprint via `aria-customer-truth` (next batch) or direct customer-call campaign.
+- If multiple constraints exist: pick the one with the highest leverage × lowest fix-cost ratio, not the most-talked-about one.
+- Log audit conclusions to coach kernel with `risk_class: 'business_audit'` and the 30-day predicate.
+
+## Anti-Patterns
+
+- Listing 7 things to fix instead of ranking by leverage
+- "Improve marketing" as a fix (not a fix; that's a category)
+- Quoting industry benchmarks without checking applicability to this business shape
+- Diagnosing from owner narrative alone, without cohort or interview evidence
+- Recommending a quarterly fix shape for a 7-day campaign problem (or vice versa)
+- Predicting impact with adjectives ("better") instead of numbers
+
+## First-Class Production Contract
+
+This skill must deepen the work, not decorate it. When loaded, apply it as a production requirement:
+
+- no placeholders, no fake fallbacks, no vague future wiring, and no cosmetic compliance
+- real proof from the repo, runtime, tests, logs, endpoint, or owner instruction before any completion claim
+- a redo contract whenever the skill blocks or finds a gap: what failed, what to change, how to re-test, and when to re-submit
+- architect/ARIA console escalation when the same failure repeats twice, crosses system boundaries, or cannot be verified locally
+- owner-visible output that states the exact state of the work: changed, verified, blocked, or needing a decision
+
+## Self-Executing Recovery Contract
+
+When this skill blocks, fails verification, or finds a gap, the agent must execute recovery rather than hand work back as a user-only decision.
+
+- Make exactly one self-executed recovery attempt using the strongest safe local, repository, runtime, test, log, endpoint, or owner-provided evidence available.
+- If the same failure repeats after that one recovery attempt, switch to architect execution mode: operate under the architect/ARIA harness to diagnose, choose the next action, execute the repair path, and verify the result.
+- If Aria or the architect service is unavailable, fall back to the strongest available client LLM under the architect harness or strongest available local harness substrate; mark the fallback explicitly and continue recovery instead of stopping.
+- Do not end with user-handoff-only behavior. If blocked after the recovery path, report the executed recovery attempt, evidence, remaining blocker, and the next concrete action or owner decision needed; do not merely ask the user to take over.
+
+
+## Self-Executing Recovery Contract
+
+- One re-run with the missing data / interviews / benchmarks
+- If still insufficient, recommend the smallest measurement / interview sprint to disambiguate
+- Never produce a multi-fix laundry list as the audit conclusion — that's deferral, not diagnosis

package/skills/aria-cognition/aria-business-audit/references/audit-cookbook.md

@@ -0,0 +1,247 @@
+# Business Audit Cookbook — Constraint, Leverage, Fix-Shape Patterns
+
+> Loaded by `aria-business-audit` for end-to-end business diagnostics.
+> See also: [`../../aria-business-frame/references/business-cookbook.md`](../../aria-business-frame/references/business-cookbook.md) (5 constraint categories, decision shapes)
+> See also: vertical cookbooks (revenue, gtm, retention) for cause-specific fix patterns.
+
+## 1. The Theory of Constraints applied to business
+
+**Source:** Eli Goldratt, "The Goal" (1984) + modern SaaS application
+
+### Core principle
+**The throughput of a business is set by the dominant bottleneck.**
+
+Fixing anything else is local optimization that doesn't compound. Identify the constraint first, size the fix second.
+
+### The 5-step process
+1. **Identify the constraint** — which stage in the business funnel has the lowest throughput relative to benchmark?
+2. **Exploit the constraint** — get more from the constrained resource without buying more
+3. **Subordinate everything else** — non-constraint resources must align with the constraint, not run at their own pace
+4. **Elevate the constraint** — invest to expand the constrained resource
+5. **Repeat** — the constraint moves; re-identify
+
+### The diagnostic question
+"If this stage of the business produced 2× as much output, would the company's ARR / cash / runway change meaningfully?"
+- Yes → this is the constraint
+- No → optimization here is wasted effort
+
+## 2. The 5-Stage Funnel Skeleton
+
+```
+ACQUISITION → ACTIVATION → RETENTION → EXPANSION → REFERRAL
+```
+
+Each stage has its own benchmarks. Map current numbers per stage, then identify the lowest-relative-to-benchmark stage.
+
+### Acquisition benchmarks (per [SaaS Metrics Benchmark 2025](https://www.rockingweb.com.au/saas-metrics-benchmark-report-2025))
+- Website→signup: 6-9% (freemium), 3-4% (trial)
+- LTV:CAC by channel (see GTM cookbook §1)
+- CAC payback: 12-24 months acceptable, <12 excellent
+
+### Activation benchmarks
+- Signup→activation: 20-40% OK, 50%+ excellent
+- Aha moment within 10 minutes of signup ideal
+- Day-7 retention of activated users: ≥60%
+
+### Retention benchmarks
+- GRR: ≥90% healthy, ≥95% best-in-class
+- NRR: 101% median 2025, 107% top quartile
+
+### Expansion benchmarks
+- NRR > GRR by ≥10 percentage points = healthy expansion
+- Multi-product attach rate: ≥30% within 12 months for cross-sell-shaped products
+
+### Referral benchmarks
+- NPS ≥50 = referral-capable
+- Customer-driven acquisition share: 10-30% = healthy network density
+- Referral conversion: 30-50% (typically 5-10× cold)
+
+## 3. The Constraint Identification Rubric
+
+For each stage, compute three numbers:
+
+1. **Current performance** (the literal number)
+2. **Benchmark performance** (what comparable companies achieve)
+3. **Leverage** (revenue / runway impact if fixed to benchmark)
+
+The stage with the **highest leverage gap** is the constraint.
+
+### Common constraint patterns by ARR stage
+
+| ARR stage | Most common constraint |
+|---|---|
+| <$1M | Acquisition (no one knows about us) OR product-market fit (we don't have lovers yet) |
+| $1M-$5M | Activation (signups don't activate fast enough) |
+| $5M-$20M | Retention (churn caps growth despite acquisition working) |
+| $20M-$50M | Expansion (we land but don't expand within accounts) |
+| $50M+ | New segment / new geo (saturating original ICP) |
+
+These are starting hypotheses, not certainties. Always verify with data.
+
+## 4. The Fix-Shape Match (campaign / system / pivot)
+
+| Constraint shape | Fix shape | Time | Cost |
+|---|---|---|---|
+| Single tactical gap (e.g., one funnel step) | Campaign | 1-3 days | Low |
+| Multi-step process (e.g., onboarding flow, sales playbook) | System | 4-12 weeks | Medium |
+| Strategic positioning (e.g., wrong ICP, wrong category) | Pivot | Quarters | High, mostly irreversible |
+
+### Mismatch failures
+- **Campaign trying to fix a system**: A/B testing the email subject line when the actual problem is the activation flow. Won't compound.
+- **System trying to fix a pivot**: rebuilding onboarding when the real problem is targeting the wrong ICP. Won't compound.
+- **Pivot when a campaign would do**: reorganizing the company when a $2K Reddit ad test would reveal the channel works.
+
+### The owner-cost asymmetry
+- A misjudged campaign costs days
+- A misjudged system costs months
+- A misjudged pivot costs years
+- Diagnose the shape carefully BEFORE recommending
+
+## 5. The Owner-Narrative vs Data Diagnostic
+
+The most common audit failure: trusting the owner narrative without verifying with data.
+
+### The pattern
+Owner says: "Our churn is bad."
+- Data check: is churn high relative to benchmark, or is the owner reading anecdotes?
+- Cohort check: is churn elevated in a specific cohort (cohort-decomposed) or blended?
+- Cause check: which of the 5 churn causes (retention cookbook §1) is dominant?
+
+Owner narrative is a hypothesis, not a finding. Verify with data before accepting.
+
+### When owner narrative IS the data
+- Customer interviews the owner did personally (qualitative data is data)
+- Pattern recognition across hundreds of conversations (gut shaped by exposure)
+- Information that hasn't yet hit the dashboards (live customer escalations)
+
+The skill: distinguish between "owner has information data doesn't yet" vs "owner has a hypothesis that data doesn't support."
+
+## 6. The 30/60/90 Plan Shape
+
+Every audit conclusion produces a 30/60/90 with explicit kill-criteria.
+
+### Template
+```markdown
+## 30 days
+- [ ] Concrete deliverable 1 (named owner, named output)
+- [ ] Concrete deliverable 2
+- [ ] Measurable signal: [specific metric ≥ threshold]
+- **Kill-criterion**: if [signal] is not met by day 30, abandon and re-diagnose
+
+## 60 days
+- [ ] Concrete deliverable
+- [ ] Measurable signal escalation
+
+## 90 days
+- [ ] Constraint moved (the original gap closed by ≥X%)
+- [ ] Measurable signal: [specific metric]
+```
+
+### The kill-criterion is mandatory
+A plan without a kill-criterion is a sunk-cost plan. The kill-criterion forces the team to face reality at the 30-day check.
+
+## 7. Benchmarks Are Conversation-Starters, Not Pass/Fail
+
+**Source:** [Fiscallion — SaaS Unit Economics](https://www.fiscallion.io/blog/saas-unit-economics)
+
+The 3:1 LTV:CAC, the 40% Rule of 40, the 90% GRR — these are starting points for diagnosis, not pass/fail grades.
+
+### The contextual diagnostic
+- **2.5:1 LTV:CAC with 9-month payback + 120% NRR** > **4:1 with 36-month payback + 95% NRR**
+- **Rule of 30 with 80% GRR** > **Rule of 50 with 60% GRR** at any growth stage
+- **Slow growth in expensive segment** > **fast growth in unsustainable segment**
+
+The audit job: name the WHY behind the number, not just the number.
+
+## 8. The 7-Powers Defensibility Audit
+
+**Source:** [Inferential Investor — 7 Powers](https://www.inferentialinvestor.com/p/the-business-moat-cheat-code-how) · See aria-business-frame business-cookbook §1
+
+For any business audit at $5M+ ARR, score the 7 Powers:
+
+| Power | Score | Evidence |
+|---|---|---|
+| Scale economies | 0-10 | [evidence] |
+| Network effects | 0-10 | [evidence] |
+| Counter-positioning | 0-10 | [evidence] |
+| Switching costs | 0-10 | [evidence] |
+| Branding | 0-10 | [evidence] |
+| Cornered resource | 0-10 | [evidence] |
+| Process power | 0-10 | [evidence] |
+
+Total /70:
+- <15: commodity / no moat → audit recommendation: build defensibility BEFORE scaling acquisition
+- 15-25: narrow moat → audit recommendation: identify the load-bearing power; double down on it
+- 26-40: investable zone → audit recommendation: scale; the moat is real
+- >40: legendary → audit recommendation: protect the moat; saturate adjacent markets
+
+## 9. The Audit Output Layout
+
+```markdown
+## [Diagnosis in one line — e.g. "Activation is the constraint; day-7 first-action drop is killing CAC payback"]
+
+### Funnel skeleton (current)
+| Stage | Current | Benchmark | Gap |
+|---|---|---|---|
+| Acquisition | 4.2% | 6-9% | -2 pts |
+| Activation | 18% | 30%+ | -12 pts ← constraint |
+| Retention | 92% GRR | 90%+ | OK |
+| Expansion | 105% NRR | 110%+ | -5 pts |
+| Referral | 8% | 15%+ | -7 pts |
+
+### The constraint, with evidence
+- **Activation** at 18% (cohort-decomposed; consistent across last 90 days)
+- Benchmark: 30%+ for [comparable shape]
+- Source: [analytics export, dated]
+
+### Leverage if fixed
+- Activation 18% → 30% = +67% activated cohort = ~$X ARR over 12 months at current acquisition rate
+- CAC payback drops from 24 months to 14 months
+- Cash impact: extends runway by ~Y months
+
+### Fix shape + 30/60/90
+- **Shape**: System (4-12 weeks)
+- **Why this shape**: not a single-tactic gap; the entire post-signup flow needs redesign
+- **30 days**: redesigned onboarding flow shipped; A/B vs control; signal: 7-day activation +5pts
+- **60 days**: full rollout; signal: blended activation +10pts
+- **90 days**: activation at 28%+; signal: CAC payback < 18 months
+- **Kill-criterion**: if 30-day signal not met, re-diagnose (constraint may be elsewhere)
+
+**Next:** approve the activation system redesign; first concrete action: 5-call user-research sprint to identify drop-off points.
+```
+
+## 10. Audit anti-patterns
+
+### Listing 7 things to fix instead of ranking by leverage
+The audit is the rank, not the list. Lists are noise.
+
+### Quoting industry benchmarks without checking applicability
+"Stripe achieves X" doesn't mean a 50-person SaaS at the same metric is failing. Match benchmarks to comparable shape.
+
+### Diagnosing from owner narrative alone
+Verify with cohort data, customer interviews, funnel measurement. Owner narrative is a hypothesis.
+
+### Recommending quarterly fix shape for 7-day campaign problem (or vice versa)
+Match the fix shape to the constraint shape. Mis-sized fixes don't compound.
+
+### Predicting impact with adjectives
+"Better" / "improved" is not a measurement. Numeric or boolean predicate, with confidence band.
+
+### Skipping the kill-criterion
+A plan without a kill-criterion is a sunk-cost plan. Force the 30-day reality check.
+
+## 11. Audit checklist (paste-ready)
+
+```markdown
+- [ ] Funnel skeleton mapped with current numbers (named source, dated)
+- [ ] Per-stage benchmark identified (comparable shape, not generic)
+- [ ] Constraint identified (highest-leverage gap)
+- [ ] Constraint cause decomposed (per relevant cookbook §)
+- [ ] Leverage size as numeric prediction with confidence band
+- [ ] Fix shape matched to constraint shape (campaign / system / pivot)
+- [ ] 30/60/90 plan with kill-criterion
+- [ ] 7-Powers score (if $5M+ ARR)
+- [ ] Owner-narrative vs data: any contradiction surfaced and resolved
+- [ ] Coach-kernel write with `risk_class: 'business_audit'` + 30-day predicate
+- [ ] Recovery contract gateType: advisory (audit findings teach; don't block)
+```

package/skills/aria-cognition/aria-business-frame/SKILL.md

@@ -0,0 +1,138 @@
+---
+name: aria-business-frame
+description: TRIGGER automatically for any non-trivial owner-outcome task — revenue, growth, pricing, GTM, sales, marketing, retention, hiring, cash flow, unit economics, partnerships, fundraising, product strategy, customer success, ops scale, brand, positioning, churn, NPS, expansion, monetization, P&L, runway, burn, CAC, LTV, ARR, MRR, conversion, funnel, activation, distribution, defensibility, moat, valuation, deal structuring, M&A, board prep, investor update, hiring plan, comp, equity, layoff, RIF, vendor selection, contract renewal, customer escalation, refund, dispute, partnership term sheet, joint venture, channel strategy, or any decision that visibly moves an owner KPI. Composes Mizan + Ladunni-22 + Ghazali-8 + Tadabbur into a single business-cognition pass before action.
+---
+
+# Aria Business Frame
+
+The root auto-fire skill for any work where the bottleneck is owner outcome, not engineering correctness. Invoke before answering — not after — on any prompt that visibly moves a KPI, customer, dollar, or runway.
+
+## Prime Doctrine
+
+A business decision without cognition becomes a tactic chasing a metric.
+
+- Engineering cognition (Mizan/Aristotle/Noor) governs HOW to build. Business cognition governs WHAT is worth building, for WHOM, at WHAT price, against WHAT alternative.
+- Apply lenses, do not perform. If the cognition does not change the price quoted, the channel chosen, the cohort targeted, the offer structured, the contract terms, or the next concrete commitment, it did not fire.
+- Owner-outcome turns demand higher cognition than code turns. A misrouted lead, a wrong refund, a stale price, a soft churn-save — these compound. A typo in a function does not.
+- Recovery is mandatory; blocking is forbidden. If a business gate fires, teach, suggest the corrected action, and let the owner proceed. Coach kernel learns; doctrine map widens; next turn is sharper.
+
+## Trigger
+
+Fire when the user prompt contains any of the following intents (auto-detected by the skill autoload gate's business-intent branch):
+
+- **Revenue / pricing**: pricing, price, package, tier, discount, free trial, paywall, monetize, monetization, ARPU, gross margin
+- **Growth / GTM**: launch, GTM, go-to-market, channel, distribution, content, SEO, paid, organic, virality, referral, PLG, sales-led
+- **Sales pipeline**: lead, prospect, qualify, demo, close, deal, contract, proposal, MEDDIC, BANT, pipeline, forecast, quota
+- **Marketing**: campaign, brand, positioning, message, copy, hook, headline, ad, creative, landing, audience, persona, ICP
+- **Retention / CS**: churn, retention, renewal, expansion, NPS, CSAT, onboarding, activation, save, escalation, refund, win-back
+- **Finance**: P&L, cash flow, runway, burn, CAC, LTV, payback, unit economics, gross margin, contribution margin, ARR, MRR
+- **Hiring / org**: hire, fire, layoff, RIF, comp, equity, performance, promotion, reorg, span of control
+- **Strategy / partnerships**: moat, defensibility, competitor, partnership, JV, channel partner, exclusive, term sheet, M&A
+- **Operations at scale**: scale, ops cost, COGS, fulfillment, support cost, automation ROI, vendor selection, contract renewal
+
+Light keyword fire by default; `batch` mode auto-engages for launch / strategy / fundraising / org-design turns.
+
+## Mode Selection
+
+Choose the smallest mode that protects owner outcome:
+
+- `light`: routine pricing question, single lead routing, copy tweak, single comp benchmark. Run Mizan intent/evidence + one Tadabbur consequence + one 8-lens **trust** + **harm** check.
+- `qa`: campaign live-fire, deal redline, refund decision, customer escalation, vendor selection, hiring decision. Load `mizan`, `tadabbur`, `ghazali-8lens`, `predictor`, `aria-business-frame`, plus the matching vertical skill (`aria-revenue-engine`, `aria-gtm-architect`, etc.).
+- `batch`: launch readiness, fundraising deck, board update, org redesign, pricing overhaul, market expansion. Load `aria-axioms-first-principles` + `aria-cognition-batch` + `ladunni-22` + `aria-business-frame` + every relevant vertical skill. Run the full owner-outcome lens stack.
+- `fleet`: simultaneous campaign + deal + ops decisions; multi-domain owner asks. Load `aria-ledger-fleet-execution` after the batch pass.
+- `teach`: a gate fired, an anti-pattern was caught — emit the teaching, name the corrected action, log to coach kernel, let the owner proceed. **Never hard-block on business turns.**
+
+## Always-On Checks (before any owner-facing recommendation)
+
+1. **What KPI does this move?** Name the metric, the current value, the predicted post-action value.
+2. **What is the cost?** Cash, time, attention, brand, reversibility. If reversibility is low, escalate cognition.
+3. **What does the customer believe right now?** Before this action, what is the prospect/customer/employee/investor model of us? After, what shifts?
+4. **What is the alternative?** No business action is unilateral — the customer can buy nothing, buy a competitor, or wait. Name the alternative explicitly.
+5. **What is the leverage point?** 80/20 — which 20% of this decision controls 80% of the outcome? Spend cognition there, not on the rest.
+6. **What is the second-order consequence?** (Tadabbur) If this works at 10x scale, what breaks? If it fails, what do we tell the next customer?
+7. **What does Hamza already know?** (Noor recognition) — search memory before generating new analysis. Stale memory is worse than no memory.
+8. **Is this an LLM-can-decide question or an owner-must-decide question?** Cognitive humility: some calls are owner-only. State that clearly.
+
+## Composition Rule (how to invoke other skills)
+
+This skill is a composer, not a replacement. It calls into:
+
+- `mizan` — for intent/evidence/protection/quality balance on any non-trivial business decision
+- `tadabbur` — for consequence chains on any irreversible or scale-sensitive action
+- `ghazali-8lens` — for truth/harm/trust/power/reflection/context/impact/beauty validation, especially on customer-facing copy and pricing
+- `ladunni-22` — for the broad domain pass on any cross-functional decision (e.g., pricing change touches sales + marketing + CS + finance + brand)
+- `predictor` — before any commitment, public claim, or launch
+- `noor-recognition` — when the answer should be retrieved from prior decisions, not invented
+- `aria-revenue-engine` — for revenue, pricing, monetization, deal structuring
+- `aria-gtm-architect` — for launch, channel, distribution, positioning
+- `aria-retention-engine` — for churn, renewal, expansion, escalation
+- `aria-decision-mizan` — for owner-must-decide turns where the LLM should refuse to choose
+- `aria-customer-truth` — for customer research, JTBD, win/loss, voice-of-customer
+- `aria-finance-realism` — for cash, runway, unit economics, valuation
+- `aria-ops-leverage` — for scale ops, automation ROI, vendor decisions
+- `aria-org-leverage` — for hiring, comp, equity, RIF, performance
+
+If a vertical skill is missing for the intent detected, use `aria-business-frame` directly with `mizan + tadabbur + ladunni-22 + ghazali-8lens` and explicitly note the gap so the autolearn queue can capture it.
+
+## Required Workflow
+
+1. **Detect KPI**: Name the owner KPI being moved. If none, ask the owner one tight question — never invent the KPI.
+2. **Recognize first**: Search memory/decisions for the same question. If Aria/Hamza already decided this in a prior turn, surface that and ask whether the context has changed before generating fresh analysis.
+3. **Frame**: Run Mizan pre-phase. State intent, evidence threshold, protection boundary, quality bar — all four in business terms (e.g., quality bar = "does this offer survive a CFO's 5-why audit?").
+4. **Reason cross-domain**: Run `ladunni-22` on which domains this decision actually touches. A pricing change is rarely just sales — it's brand + finance + CS + ops + product all at once.
+5. **Apply lenses**: Run `ghazali-8lens` with business-mapped definitions:
+   - **truth** — is the data current? was it sampled correctly?
+   - **harm** — what could go wrong for the customer / employee / counterparty?
+   - **trust** — does this preserve or burn brand/customer/team trust?
+   - **power** — what asymmetry are we using or creating?
+   - **reflection** — would Hamza recognize this as his decision in 6 months?
+   - **context** — what changes about this answer in 30/90/365 days?
+   - **impact** — what is the predicted KPI delta + confidence?
+   - **beauty** — does the offer/copy/contract feel inevitable, or assembled?
+6. **Predict**: Run `predictor` with a measurable predicate — numeric, boolean, or named state. No qualitative "should improve."
+7. **Recommend with three legs**: Thinking (cognition shown), implementation (concrete next action), accountability (how the owner will know it worked + when).
+8. **Recover, never block**: If the lens flags concern, teach the gap, propose the corrected action, let the owner choose. Log the gap to the coach kernel.
+
+## Recovery Contract
+
+When this skill blocks, fails verification, or finds a gap:
+
+- **One self-executed recovery**: Re-run the corrected analysis with the missing lens applied. Do not return the gap as a question to the owner unless cognition genuinely cannot resolve it.
+- **Coach kernel teaching**: Write a `recordCoachPhase({ phase: 'post_cognition', decision: 'taught', risk_class: 'business_lens_gap', reasons: [...], next_action: 'corrected recommendation' })` so the next turn opens with this gap closed.
+- **Autolearn queue**: If the gap was caused by a novel pattern not in `doctrine_trigger_map.json`, append a candidate trigger entry to `~/.claude/aria-trigger-queue.jsonl` with `severity: warn` (recovery-only, never block).
+- **Owner-visible**: State exactly what was caught, what changed, and what the owner should ratify or override. No silent self-correction.
+- **No hard block on business turns**: The autoload gate runs in `redirectOnly: true` mode for business intents. Missing skills produce a teaching, not a refusal.
+
+## First-Class Production Contract
+
+This skill must deepen the work, not decorate it. When loaded, apply it as a production requirement:
+
+- no placeholders, no fake fallbacks, no vague future wiring, and no cosmetic compliance
+- real proof from the repo, runtime, tests, logs, endpoint, or owner instruction before any completion claim
+- a redo contract whenever the skill blocks or finds a gap: what failed, what to change, how to re-test, and when to re-submit
+- architect/ARIA console escalation when the same failure repeats twice, crosses system boundaries, or cannot be verified locally
+- owner-visible output that states the exact state of the work: changed, verified, blocked, or needing a decision
+
+## Self-Executing Recovery Contract
+
+When this skill blocks, fails verification, or finds a gap, the agent must execute recovery rather than hand work back as a user-only decision.
+
+- Make exactly one self-executed recovery attempt using the strongest safe local, repository, runtime, test, log, endpoint, or owner-provided evidence available.
+- If the same failure repeats after that one recovery attempt, switch to architect execution mode: operate under the architect/ARIA harness to diagnose, choose the next action, execute the repair path, and verify the result.
+- If Aria or the architect service is unavailable, fall back to the strongest available client LLM under the architect harness or strongest available local harness substrate; mark the fallback explicitly and continue recovery instead of stopping.
+- Do not end with user-handoff-only behavior. If blocked after the recovery path, report the executed recovery attempt, evidence, remaining blocker, and the next concrete action or owner decision needed; do not merely ask the user to take over.
+
+
+## Self-Executing Recovery Contract
+
+When a sister skill (mizan/8-lens/tadabbur) blocks under this composer:
+
+- Run exactly one recovery pass with the corrected frame.
+- If still blocked, switch to `aria-decision-mizan` and explicitly mark the turn as owner-must-decide. Do not paper over with confidence theatre.
+- If Aria architect/runtime is available, escalate the structural pattern (not the single decision) to the architect harness via `runWithGovernance()`.
+- If unavailable, fall back to the strongest local cognition pass under `aria-cognition-batch` and mark the fallback explicitly.
+- Never end with "what would you like me to do?" alone — always pair the question with the cognition surfacing it.
+
+## Notes on Why This Exists
+
+The repo's existing 51 cognitives govern engineering correctness with surgical precision. They were forged in the deploy/refactor/audit fire. Owner-outcome turns are equally consequential and were under-served. This skill is the bridge: same cognitive substrate (Mizan/Aristotle/Noor/Ghazali/Ladunni), business-mapped definitions, recovery-not-block posture, coach-kernel teaching loop.