@vextlabs/theron-cli 0.3.0 → 0.4.0

package/skills/lease-review.md

@@ -0,0 +1,93 @@
+---
+name: lease-review
+description: Review a commercial lease end-to-end: extract and stress-test every economic term (base rent, escalations, CAM, TI, free rent, options), flag off-market risk provisions (assignment, default cure periods, exclusivity, co-tenancy, holdover), and deliver a redline priority list with effective-rent computation.
+allowed-tools: Read, Write
+---
+
+═══ HARD RULES ═══
+
+- Never present a market-rate benchmark as fact unless the lease itself states it or the user provides comparables; flag every benchmark as "subject to local market verification."
+- Never advise on legal enforceability — flag clauses for attorney review, never opine on state-specific law.
+- Always distinguish between gross, modified-gross, NNN, and full-service-gross lease structures before quoting any dollar figure; the structure determines what the tenant actually pays.
+- Never net a TI allowance against base rent in the effective-rent calc unless the lease explicitly amortizes it into rent.
+- Flag every undefined or landlord-discretion cap as a material risk; do not assume good faith.
+- If a lease section is missing or marked "TBD," treat it as landlord-favorable and flag it as a priority gap.
+- Compute effective rent on a per-SF/per-year basis AND as total nominal dollars over the full initial term.
+
+═══ PHASE A — DOCUMENT INTAKE AND STRUCTURE MAP ═══
+
+1. Identify the lease type (NNN / modified gross / full-service gross / gross with base-year stop). All downstream math depends on this.
+2. Record the parties: landlord entity, tenant entity, guarantor(s), and whether any guaranty is personal, corporate, or capped (burn-down schedule).
+3. Extract the premises: rentable square footage (RSF), usable SF if stated, floor, suite, load factor / add-on ratio. Verify that RSF drives the rent calc.
+4. Map the critical dates: commencement, rent-commencement (after free-rent), expiration, each option-exercise deadline, each renewal/expansion window. Build a single chronological date table.
+5. Flag any landlord termination right (recapture, demolition, relocation) and note whether tenant gets notice, a termination payment, or nothing.
+
+═══ PHASE B — ECONOMIC TERM EXTRACTION ═══
+
+6. BASE RENT: Pull every rent schedule line — Year 1 base, each step-up year, and the mechanism (fixed-dollar bumps vs. percentage bumps vs. CPI). Compute the annual rent/RSF for every lease year.
+7. ESCALATIONS: Identify the escalation type.
+   - Fixed-step: verify the step size is stated as a dollar amount or percentage (not both, not vague).
+   - CPI: confirm index (CPI-U All Items / CPI-W / local variant), base month, cap on any single-year increase ("CPI but not more than X%"), and floor ("not less than Y%"). A CPI provision with no cap is tenant-adverse.
+   - Porter's Wage or index-specific: note the index and any lag.
+8. CAM / OPERATING EXPENSES: For NNN or modified-gross leases:
+   - List every category the landlord may charge to CAM (janitorial, management fee, insurance, real estate taxes, capital repairs, landscaping, parking lot).
+   - Identify exclusions the tenant negotiated (landlord's financing costs, above-grade capital, leasing commissions, depreciation). Flag any category NOT excluded as a risk.
+   - Check whether management fees are capped (standard: 3–5% of gross revenues; flag if uncapped or if charged on top of actual third-party management).
+   - Identify the CAM cap: gross-up cap (controllable expenses only, typically 3–5% cumulative or year-over-year), whether capital is excluded from the cap, and what "controllable" means in the lease.
+   - Note base-year / base-amount stop structures and verify the base year is a fully occupied, stabilized year; if not, tenant may overpay from day one.
+9. TENANT IMPROVEMENT ALLOWANCE (TI):
+   - State the TI/RSF, total dollar amount, and what it covers (hard costs only vs. soft costs, FF&E, cabling, permit fees).
+   - Note the draw process: required documentation, landlord approval rights, disbursement timing, and whether unused TI is forfeited or credited to rent.
+   - Flag any "as-is" delivery with no TI as a risk requiring an independent condition assessment.
+10. FREE RENT: Record the number of months, which months (base only vs. base + CAM vs. full abatement), and whether free rent is conditional on no default (i.e., repayment obligation if tenant defaults — standard but must be flagged).
+11. OPTIONS TO RENEW / EXPAND / TERMINATE:
+    - For each option: notice window (how far in advance, exact calendar trigger), rent-setting method (fixed, FMV, FMV with floor/ceiling), and whether the landlord has a recapture right upon exercise.
+    - FMV options: confirm whether tenant has a right to dispute the landlord's FMV determination (arbitration, baseball arbitration, ROFR on third-party offers). No dispute mechanism = landlord sets the rate unilaterally.
+    - Expansion options: confirm whether the right is a ROFO (right of first offer), ROFR (right of first refusal), or a must-take obligation. ROFR requires tenant to respond to a live deal; ROFO is proactive — flag the difference.
+    - Termination options: note any termination fee (typically equivalent to unamortized TI + unamortized leasing commission + X months of rent), the exercise window, and blackout periods.
+
+═══ PHASE C — RISK TERM ANALYSIS ═══
+
+12. ASSIGNMENT AND SUBLETTING:
+    - Confirm whether landlord consent is required and what "not to be unreasonably withheld" means (time limit on response? deemed-approved if no response?).
+    - Identify recapture rights: if landlord can recapture on an assignment or sublease request, the option right has no value if the space is hot.
+    - Check whether assignment is permitted without consent to affiliates, parent/subsidiary, or in a corporate restructuring / change of control. Change-of-control triggers are critical for M&A scenarios.
+    - Flag profit-sharing provisions on subleases (landlord takes 50–100% of any excess rent above tenant's base — negotiate to net out TI amortization and sublease costs first).
+13. DEFAULT AND CURE PERIODS:
+    - Monetary default: how many days to cure after written notice? Standard range: 5–10 business days for first instance, extended cure for subsequent.
+    - Non-monetary default: how many days to cure? Is there a "reasonable time to cure if diligently pursuing" carve-out for complex defaults (construction, permits)?
+    - Cross-default: does default under any other landlord lease or guarantor obligation trigger default here? Flag.
+    - Self-help rights: if landlord can enter and cure at tenant's expense after notice, verify cure-cost reimbursement is limited to reasonable documented costs, not landlord's inflated estimate.
+14. EXCLUSIVITY:
+    - State the protected use category verbatim; narrow language protects nobody.
+    - Confirm the radius of protection (is it co-tenants only, or the entire project / adjacent landlord-owned parcels?).
+    - Identify carve-outs (existing tenants, anchor leases, incidental sales). Incidental-sales carve-outs above 10–15% of a co-tenant's revenue are routinely abused.
+    - State the remedy for violation: rent abatement, termination right, or damages only. Damages-only exclusivity is unenforceable in practice.
+15. CO-TENANCY:
+    - Opening co-tenancy: is rent commencement or lease commencement conditioned on anchor(s) opening? What is the remedy if the condition is not met (delayed commencement, reduced rent, or walk right)?
+    - Ongoing co-tenancy: if a named anchor or a minimum-occupancy threshold (e.g., 75% of GLA occupied and open) fails, what is the remedy? Reduced percentage/flat rent during the failure period, and a walk right if the failure persists beyond a stated cure period (12–18 months is standard).
+    - Confirm that "open and operating" is defined — a dark anchor that holds its lease defeats co-tenancy without a definition.
+16. HOLDOVER:
+    - State the holdover rent rate (common: 150% of the last month's base rent, sometimes 200%). Confirm whether it is base rent only or includes CAM.
+    - Note whether holdover converts to month-to-month or creates a tenancy at sufferance (landlord can terminate immediately). Month-to-month holdover is significantly more tenant-friendly.
+    - Flag any consequential-damages exposure if holdover causes landlord to lose a successor tenant — this is non-standard and should be negotiated out.
+
+═══ PHASE D — EFFECTIVE RENT COMPUTATION ═══
+
+17. Collect inputs: base rent schedule by year, estimated CAM (use landlord's estimate if no actuals; note the source), free-rent months, TI allowance, any landlord work letter credit.
+18. Compute GROSS RENT (base + estimated CAM) for each lease year on a $/RSF/yr basis.
+19. Apply free-rent abatement: convert free-rent months to a per-year credit spread across the full initial term.
+20. Compute NET EFFECTIVE RENT = (Total nominal rent over initial term − free-rent value) ÷ (RSF × lease years).
+21. If TI is material, compute the LANDLORD EFFECTIVE COST and present the all-in economics from both sides: tenant's net effective rent and landlord's effective yield after TI and free rent.
+22. Present the result as: (a) $/RSF/yr net effective rent, (b) total nominal obligation over the initial term, (c) average annual cash outlay including estimated CAM. Label every estimated input as estimated.
+
+═══ PHASE E — REDLINE PRIORITY OUTPUT ═══
+
+23. Produce a tiered issue list:
+    - TIER 1 — DEAL-BREAKERS: terms that expose tenant to unbounded liability or eliminate a negotiated right (uncapped CAM, recapture on assignment, no FMV dispute mechanism, consequential holdover damages).
+    - TIER 2 — MATERIAL RISKS: terms that are landlord-favorable but remediable by negotiation (short cure periods, no change-of-control carve-out, narrow exclusivity remedy, undefined CAM controllable cap).
+    - TIER 3 — MARKET CLEANUP: standard items often left in first-draft landlord forms that experienced tenants routinely clean (management fee cap, gross-up base year, TI draw documentation requirements, sublease profit-sharing formula).
+24. For each Tier 1 and Tier 2 item, propose the specific language or structural change the tenant should request — not just the concept, but the redline framing.
+25. Close with a one-paragraph EXECUTIVE SUMMARY: lease type, initial term, net effective rent/RSF/yr, top three risks, and a go/no-go signal (Proceed / Proceed with Redlines / Renegotiate Before Proceeding).
+
+KEY PRINCIPLE: Every dollar the landlord gives back (free rent, TI, options) is worth nothing if the risk clauses let them take it all back — price and risk are one underwriting decision, not two.

package/skills/lesson-plan.md

@@ -0,0 +1,198 @@
+---
+name: lesson-plan
+description: Build a rigorous, topic-specific lesson from scratch — diagnose entry state, write Bloom-leveled objectives, sequence concrete-to-abstract, scaffold worked examples into faded practice, and preempt the misconceptions that derail learners of this exact topic.
+allowed-tools: Read, Write
+---
+
+## ═══ HARD RULES ═══
+
+1. Every objective must name a MEASURABLE BEHAVIOR at a stated Bloom level — "understand" is never acceptable; "construct," "distinguish," "critique," "predict," "derive," "execute" are.
+2. Concrete before abstract, ALWAYS. No definition precedes the example that makes it necessary. No formula precedes the situation that demands it.
+3. Never fabricate citations, statistics, or benchmark figures as ground truth. If you need a number to illustrate a principle, label it explicitly as illustrative.
+4. Worked examples are NOT optional decoration — include at least two fully worked examples per enabling objective before reducing scaffolding. Partial worked examples do not count.
+5. Formative checks must be diagnostically useful: each wrong-answer option must correspond to a specific, named misconception or error type. A check that reveals only "right or wrong" is not a formative check.
+6. Prerequisite gaps discovered during the session PAUSE the lesson — you do not advance the sequence over a broken foundation under any circumstances.
+7. Fading is a property of the scaffold, not the difficulty. Reduce worked support progressively; do not reduce the cognitive challenge of the target material.
+8. Misconception preemption belongs BEFORE the point where the misconception forms, not in a remediation appendix after damage is done.
+9. One lesson = one COHERENT conceptual move. If the lesson contains two independent conceptual moves, split it into two lessons.
+10. Coverage is not learning. Covering material a learner cannot yet process is waste.
+11. SAFETY: Do not diagnose learning disabilities, psychological conditions, or neurodevelopmental profiles. Do not produce instructional content for high-stakes licensed or credentialed domains (medicine, law, aviation) without flagging that a qualified human instructor must review it. Accuracy of domain-specific claims is the author's responsibility — this skill governs instructional design, not subject-matter authority.
+
+---
+
+## ═══ PHASE A — INTAKE: SPECIFY THE LEARNING TARGET ═══
+
+1. State the topic at the GRAIN SIZE of a single lesson — specific enough that a competent practitioner could teach it in one sitting. Reject vague scope:
+   - Too broad: "calculus" / "writing" / "machine learning"
+   - Correct grain: "Applying integration by parts to products of polynomial and exponential functions" / "Writing a thesis statement that makes a contestable claim" / "Choosing between L1 and L2 regularization given a feature distribution"
+
+2. Name the target learner population with four parameters:
+   - **Prior exposure**: what they have already learned and when
+   - **Domain fluency**: how comfortable they are with the notation, vocabulary, and conventions of this field
+   - **Motivation**: why they are learning this (job requirement, exam, intrinsic interest, remedial catch-up)
+   - **Context of use**: classroom, self-study, on-the-job, asynchronous — this determines pacing and feedback loops
+
+3. Write the terminal performance objective in this exact form:
+   > "Given [specific stimulus], the learner will [observable verb at target Bloom level] [specific object], such that [measurable quality criterion]."
+
+   Bloom verb anchors by level:
+   - **Remember**: recall, list, name, identify
+   - **Understand**: explain, paraphrase, classify, summarize
+   - **Apply**: solve, compute, use, execute, demonstrate
+   - **Analyze**: differentiate, decompose, compare, attribute
+   - **Evaluate**: justify, critique, judge, defend, prioritize
+   - **Create**: construct, design, generate, produce, formulate
+
+4. Decompose the terminal objective into 3–6 enabling objectives, each at Bloom-Apply or below — the sub-skills jointly sufficient for the terminal. Number them (EO-1 through EO-N); they become the lesson spine. Reject enabling objectives that are restatements of the terminal objective at lower resolution.
+
+---
+
+## ═══ PHASE B — PREREQUISITE DIAGNOSIS ═══
+
+5. For each enabling objective, name the ONE prerequisite concept or skill it directly depends on. Build the prerequisite tree. If an enabling objective has no meaningful prerequisite, state that explicitly — do not fabricate one.
+
+6. Write a 2–4 item diagnostic probe per prerequisite. Each probe must:
+   - Be answerable in under 90 seconds
+   - Test the prerequisite directly, not the terminal skill
+   - Have an unambiguous correct response the practitioner can evaluate without domain expertise in the full topic
+   Administer or present these probes BEFORE any instruction begins.
+
+7. Interpret probe responses and branch:
+   - **All correct**: proceed to Phase C.
+   - **Specific gap found**: provide targeted bridging explanation for that gap only; re-probe with a parallel item; proceed only on success.
+   - **Pervasive gaps**: surface this explicitly — the current lesson requires prior lessons first. Name which prerequisite lesson is missing. Do not paper over a broken foundation with extra scaffolding in the current lesson.
+
+8. Record the learner's entry state in a one-sentence profile before Phase C begins. This profile drives every density, pacing, and sequencing decision downstream.
+
+---
+
+## ═══ PHASE C — SEQUENCE DESIGN: CONCRETE TO ABSTRACT ═══
+
+9. For each enabling objective, identify a CONCRETE ANCHOR — a real, specific, observable instance of the concept in use within the learner's existing experience. The anchor must be:
+   - Familiar to this specific learner population (not to you)
+   - Unambiguous: any two practitioners looking at it would agree it is an instance of the concept
+   - Minimal: it should carry the concept without excess detail that distracts from structure
+
+   Pattern for each enabling objective:
+   - Concrete anchor (real instance) → Observations about the anchor → Abstract pattern extracted → Formal definition or rule → Boundary case showing where the rule fails
+
+10. Sequence enabling objectives by dependency, not by textbook chapter order, alphabetical convenience, or apparent simplicity. A learner must be able to perform EO-N before EO-(N+1) is even meaningful.
+
+11. Identify the CONCEPTUAL PIVOT — the single moment in the sequence where the learner transitions from recognizing examples to generating them. This is the highest cognitive load point in the lesson. Design the densest scaffolding here, not at the beginning.
+
+12. Draft a practitioner script for each enabling objective with these four beats in sequence:
+    - **Hook** (1–3 sentences): a question or situation that creates the felt need for this concept — cognitive dissonance or a puzzle, not a trivia question or motivational preamble
+    - **Anchor example** (fully narrated): a worked-through, concrete instance with decision narration (see Phase D)
+    - **Pattern extraction** (explicit, spoken aloud): the generalizable structure, stated in plain language before any formal notation
+    - **Boundary** (one non-example): a case that looks like it should work but does not, showing where the concept's valid domain ends
+
+    The script is a practitioner-facing planning tool, not a learner-facing transcript. Write it in the second person ("You will show...") so the practitioner knows what to do.
+
+---
+
+## ═══ PHASE D — WORKED EXAMPLES ═══
+
+13. Write at least TWO fully worked examples per enabling objective. Both must be worked to completion — no steps elided, no "it can be shown that," no "the rest follows similarly."
+
+14. Narrate each step from the DECISION standpoint, not the operation standpoint:
+    - **Operational narration** (insufficient): "Multiply both sides by 2."
+    - **Decision narration** (required): "We multiply both sides by 2 because we need to isolate x and this is the inverse of dividing by 2 — not because 2 appears in the problem."
+
+15. Vary the SURFACE FORM between worked examples while holding the underlying structure constant. If EO-1 concerns a rate-of-change interpretation, work one example with velocity and one with population growth — same schema, different context. This forces attention to structure.
+
+16. Include one NEAR-MISS worked example per enabling objective — a plausible attempt that makes the most common error for this topic, followed by explicit diagnosis naming the error and a correction that shows where the reasoning diverged. Near-miss examples are more effective for misconception preemption than a second correct example.
+
+17. Worked examples precede ANY practice attempt. "Try it first and see what happens" is appropriate for exploration; it is not appropriate for novices acquiring a new schema.
+
+---
+
+## ═══ PHASE E — FADED PRACTICE ═══
+
+18. Fading means reducing scaffolding across a problem set, not increasing problem difficulty. Hold the conceptual level constant; remove progressively more of the worked solution at each step.
+
+    Fade sequence (four mandatory steps):
+    - **Completion problems**: problem + first 60–70% of the solution worked out; learner supplies the remainder
+    - **Partial problems**: problem + first 30–40% worked; learner generates the rest
+    - **Prompted problems**: problem + a structural prompt specifying the first decision (e.g., "What do you need to isolate, and why?"); no worked content
+    - **Independent problems**: problem only
+
+19. Each faded step is a SEPARATE ITEM. Learners do not self-select their scaffold level — the practitioner controls the fade sequence.
+
+20. After every completion at a fade level, require the learner to EXPLAIN the step they just completed — written or spoken. The self-explanation requirement is not optional and is not satisfied by "yes, that's right." The learner must articulate the decision, not confirm correctness.
+
+---
+
+## ═══ PHASE F — INDEPENDENT PRACTICE ═══
+
+21. Independent practice items must require GENERATION, not recognition. Multiple-choice is a diagnostic instrument at this stage; it is not a practice instrument.
+
+22. Build the practice set to vary along exactly TWO dimensions, independently:
+    - **Surface context**: different story, domain, notation, or numbers — same underlying structure
+    - **Structural variant**: isomorphic structure in a different configuration — same concept, different arrangement
+    Do not conflate surface variation with structural variation and call it "different problems."
+
+23. INTERLEAVE enabling objectives within the practice set — do not block all items for EO-1, then all for EO-2. Interleaving forces the learner to discriminate which schema applies, producing substantially better long-term retention than blocking, at the cost of feeling harder in the moment. Tell the learner this explicitly so they do not interpret difficulty as failure.
+
+24. Include at least one TRANSFER item — a problem where the learned schema applies but is not labeled as belonging to the current lesson topic. Transfer items reveal whether the learner has encoded a surface pattern or a structural understanding.
+
+25. Do not assign more independent practice than is necessary to confirm fluency at the terminal objective. Excess practice generates compliance fatigue and surface responding, not deeper encoding.
+
+---
+
+## ═══ PHASE G — FORMATIVE CHECKS ═══
+
+26. Place a formative check at the END of each enabling objective — before advancing to the next. A check placed only at the end of the full lesson is a summative assessment, not a formative check.
+
+27. Each formative check item must be DIAGNOSTIC. Structure:
+    - **Stem**: a minimal, unambiguous problem at the enabling objective level
+    - **Correct answer**: the target behavior for this objective
+    - **Distractor 1**: the result of the most common procedural error for this specific topic (name the error explicitly in your planning notes)
+    - **Distractor 2**: the result of the most common conceptual error for this specific topic (name the misconception explicitly)
+    - **Distractor 3 (if warranted)**: the result of the most likely prerequisite failure
+
+    If you cannot name what a distractor reveals, replace it.
+
+28. Use the check response to BRANCH immediately:
+    - **Correct**: advance to the next enabling objective.
+    - **Distractor 1 selected**: deliver procedural correction — re-narrate the specific decision at the failing step; re-probe with a parallel item.
+    - **Distractor 2 selected**: deliver conceptual correction — return to the concrete anchor; re-extract the pattern explicitly; re-probe.
+    - **Distractor 3 selected**: return to the prerequisite and address it before re-entering the lesson sequence.
+
+29. Log which misconceptions are active for this learner after each check. Misconceptions compound — an unresolved misconception at EO-2 will corrupt performance at EO-4 and EO-5 in ways that are hard to trace back to source.
+
+---
+
+## ═══ PHASE H — MISCONCEPTION PREEMPTION ═══
+
+30. Before writing any instructional content, produce a MISCONCEPTION INVENTORY for this specific topic. Method:
+    - List every wrong belief a learner could plausibly hold about this exact concept, given the stated prior exposure of the target population
+    - For each: state (a) the surface symptom — what the learner says, writes, or does that reveals the belief; (b) the root belief that generates the symptom; (c) why the root belief is locally coherent (it usually worked somewhere else before failing here)
+    - Order by estimated frequency in the target population, not by severity
+
+31. For each misconception in the inventory, write ONE targeted counter-example that creates productive cognitive conflict — a case where the misconception predicts outcome X and the correct reasoning predicts outcome Y, and Y is verifiably true. The counter-example must be minimal: no extraneous variables that let the learner escape by attributing the conflict to something else.
+
+32. Embed each counter-example at the point in the lesson sequence JUST BEFORE the misconception is most likely to form. This is always before first exposure to the concept, not after practice reveals the error.
+
+33. To identify the misconceptions specific to your topic (do not import generic lists), use this elicitation protocol:
+    - **Prior-knowledge probe**: what would a learner with the stated entry state predict before instruction? The gap between prediction and correct answer reveals the active prior belief.
+    - **Common error audit**: what errors does the target population actually produce in assessments? Each error pattern traces back to a belief.
+    - **Analogy trap check**: what concepts or rules from prior learning look similar enough to this topic that a learner would incorrectly apply them here? (e.g., a rule that works in one domain being over-extended into this one)
+    - **Structural category**: once you have identified the specific misconception, classify it — overgeneralization, undergeneralization, conflation of two distinct concepts, inversion of a causal or directional relationship, or correct concept with misapplied procedure. The category determines the correction strategy.
+
+34. After the lesson, retrospectively audit which misconceptions appeared in formative check responses. Record frequency data. Update the misconception inventory for future instruction on this topic.
+
+---
+
+## ═══ PHASE I — LESSON CLOSURE AND TRANSFER PREP ═══
+
+35. Close the lesson with a CONSOLIDATION move — the learner generates the abstract pattern from memory, in their own words, without scaffolding. This is not a summary you provide; it is a retrieval attempt they produce. Do not confirm or correct until they have committed to a full response.
+
+36. Pose one TRANSFER PROMPT — a novel context where the learned schema applies and the learner must recognize it is relevant without being told so. The context should be meaningfully different from all practice items used in the lesson.
+
+37. State explicitly what the next lesson will build on from this one. Make the dependency visible — the learner should be able to name what they just earned access to.
+
+38. Identify the SPACING recommendation: name a specific time window for first review of this material. The spacing effect is well-documented but optimal intervals vary by topic complexity, prior exposure, and desired retention duration; 24–48 hours post-acquisition is a reasonable illustrative default for novel conceptual content, not a universal prescription. State the recommendation as a specific calendar target, not as "soon."
+
+---
+
+KEY PRINCIPLE: A lesson is not a presentation of content — it is a carefully sequenced series of cognitive acts performed BY the learner, arranged so that each act builds the next and no act is attempted before its foundation is in place. The practitioner's job is to engineer that sequence, not to deliver information.

package/skills/literature-review.md

@@ -0,0 +1,69 @@
+---
+name: literature-review
+description: Systematic literature review — search strategy, screening, quality appraisal, thematic synthesis, and an honest gap analysis with primary-source citations.
+allowed-tools: WebSearch, WebFetch, Read, Write, Grep, TodoWrite
+---
+
+## PHASE 0 — FRAME THE QUESTION (do this before any search)
+
+1. State the question in PICO/PECO form: **P**opulation/Problem, **I**ntervention or **E**xposure, **C**omparison, **O**utcome. If non-clinical, map the analogous axes. Write it out explicitly — every subsequent decision derives from it.
+2. Declare scope constraints up front: date range, language, study design eligibility, minimum sample size (if any), geographic scope.
+3. Use TodoWrite to create a checklist of all phases below before starting. Track progress there; do not skip ahead.
+
+## PHASE 1 — SEARCH STRATEGY (must be stated explicitly, not implied)
+
+4. Select databases appropriate to the domain: PubMed/MEDLINE, Semantic Scholar, arXiv, ACM DL, IEEE Xplore, SSRN, Cochrane, Web of Science, Google Scholar as fallback. State which are used and why.
+5. Construct boolean queries: combine MeSH/controlled vocabulary terms OR free-text synonyms with AND/OR/NOT. Write the exact query string for each database before running it.
+6. Run WebSearch with those exact strings. For each result page, use WebFetch on the landing page or abstract to extract title, authors, year, venue, DOI/URL.
+7. Record every search string used and the hit count returned. Do not silently drop queries that return few results — note them.
+8. Search for grey literature where relevant: preprint servers, government/agency reports, trial registries (ClinicalTrials.gov, WHO ICTRP). Flag these as grey literature in the extraction matrix.
+9. Check reference lists of the 3-5 most relevant papers found (backward citation chaining) and the citing papers of landmark works (forward chaining via Semantic Scholar "Cited By").
+
+## PHASE 2 — SCREENING (two-pass, explicit criteria)
+
+10. Define inclusion criteria before screening: study design(s) accepted, population match, intervention/exposure match, outcome measured, minimum follow-up if applicable.
+11. Define exclusion criteria: editorials/opinion pieces, studies without primary data, sample size below threshold, off-topic outcomes, duplicates.
+12. Pass 1 — title/abstract screen: apply criteria to each retrieved record. Mark each as INCLUDE / EXCLUDE / UNCERTAIN. Do not resolve UNCERTAIN by guessing — fetch the full text.
+13. Pass 2 — full-text screen: WebFetch the full text (or abstract + methods section) for all INCLUDE and UNCERTAIN. Apply criteria again. Record the exclusion reason for every excluded paper.
+14. Deduplicate: if the same study appears under multiple records (preprint + journal), keep the most recent peer-reviewed version and note the duplicate.
+
+## PHASE 3 — QUALITY APPRAISAL
+
+15. Apply the appropriate tool to each included study:
+    - RCTs: Cochrane RoB 2 (randomisation, blinding, attrition, selective reporting).
+    - Observational: ROBINS-I or Newcastle-Ottawa Scale.
+    - Diagnostic accuracy: QUADAS-2.
+    - ML/AI papers: report dataset size, train/test split, leakage risk, benchmark contamination, external validation presence.
+    - Mechanistic/theoretical: assess logical soundness, whether claims are empirically tested or conjectured.
+16. Assign each paper a quality tier: HIGH / MODERATE / LOW / VERY LOW (GRADE-aligned where applicable). State the dominant reason for downgrading.
+17. Never promote a single-study finding to "established" regardless of journal prestige.
+
+## PHASE 4 — DATA EXTRACTION MATRIX
+
+18. For every included paper, extract into a row: Author(Year) | DOI/URL | Study design | N (sample size) | Population | Intervention/Exposure | Comparator | Primary outcome + effect size (with 95% CI) | Follow-up | Quality tier | Key limitation.
+19. Effect sizes must be verbatim from the paper — do not recompute or round silently. If unreported, write "NR" and flag as a limitation.
+20. Separate quantitative outcomes (mean difference, OR, RR, HR, AUC, accuracy %) from qualitative findings. Do not mix them in the same synthesis cell.
+
+## PHASE 5 — THEMATIC SYNTHESIS (never paper-by-paper summary)
+
+21. Group papers by theme/mechanism/outcome cluster, not by publication order. Identify 3-7 themes from the PICO outcome axis and the patterns in the extraction matrix.
+22. For each theme: summarize the weight of evidence (N studies, total N participants), the direction and consistency of effects, the magnitude range (min–max effect size across studies), and the quality tier distribution.
+23. Use explicit epistemic labels on every claim:
+    - **ESTABLISHED**: consistent direction across ≥3 independent high/moderate-quality studies with no serious contradicting evidence.
+    - **CONTESTED**: studies exist on both sides, or heterogeneity (I² > 60% if meta-analytic) is unexplained.
+    - **SPECULATIVE**: single study, very low quality, or theoretical only — no replicated empirical support.
+24. Where a meta-analysis exists and is methodologically sound, report its pooled estimate. Do not pool numbers yourself unless you can show homogeneity.
+25. Identify moderators and subgroup effects reported in primary studies; note which are pre-specified vs post-hoc.
+
+## PHASE 6 — GAP ANALYSIS
+
+26. State explicitly what the literature does NOT answer relative to the framed PICO question. Distinguish: (a) gaps due to no study existing, (b) gaps due to methodological limitations in existing studies, (c) gaps due to population/setting exclusions.
+27. Identify the highest-priority unanswered question and, where possible, sketch what study design would address it.
+
+## PHASE 7 — REPORTING RULES (hard constraints, no exceptions)
+
+28. **Primary sources only.** Cite the original study, not a review that cited it. If you only have access to a secondary description, say so explicitly and do not present it as a verified primary-source fact.
+29. **Never fabricate a citation, author, title, DOI, or finding.** If a paper cannot be verified via WebSearch or WebFetch, it does not appear in the review. Write "could not verify" and omit.
+30. **Flag provenance on every factual claim**: append `[WebSearch]`, `[WebFetch: <url>]`, or `[prior knowledge — unverified]`. Prior-knowledge claims must be corroborated by a live fetch before being presented as established.
+31. **Report limitations of this review itself**: databases searched, date cutoff, language bias, inability to access paywalled full texts, time constraints that prevented exhaustive screening.
+32. **Do not overstate**: if evidence is thin, say so. Honest uncertainty is a feature, not a weakness.

package/skills/market-entry.md

@@ -0,0 +1,137 @@
+---
+name: market-entry
+description: Analyze a new market entry decision end-to-end: market attractiveness, right-to-win, entry mode and sequencing, irreversible vs reversible commitments, strongest reasons NOT to enter, and a staged de-risking plan with explicit kill criteria.
+allowed-tools: Read, WebSearch, Bash, Write
+---
+
+═══ HARD RULES ═══
+
+1. NEVER assert a market size, CAGR, or competitor share as ground truth without citing a primary or traceable secondary source found this session; mark estimates as ESTIMATE with your confidence band.
+2. NEVER recommend entry without completing ALL phases; skipping a phase to reach a conclusion faster is a disqualifying error.
+3. NEVER conflate "market is large" with "we can win a piece of it" — attractiveness and right-to-win are independent questions with independent evidence burdens.
+4. NEVER omit the anti-thesis section (Phase E); a case with no serious reasons NOT to enter has not been stress-tested.
+5. NEVER recommend an irreversible commitment (factory, exclusive license, M&A) as a Stage-1 move without explicit board-level sign-off noted in the output.
+6. ALL kill criteria must be falsifiable, time-bound, and metric-specific — no fuzzy gates like "traction is insufficient."
+7. Entry mode selection must name the specific legal/commercial vehicle (e.g., wholly-owned subsidiary, JV with minority stake, white-label OEM agreement, distributor agreement, asset acquisition) — generic labels are disallowed.
+8. NEVER present a market as a simple product market without first checking for platform dynamics (multi-sided network effects, winner-take-most concentration). Platform markets require a fundamentally different entry calculus and must be flagged immediately if present.
+9. NEVER skip the timing question: "why now?" must be answered with a structural market window argument, not a statement of the entrant's readiness.
+
+═══ PHASE A — MARKET DEFINITION, SCOPING, AND TIMING ═══
+
+A1. State the precise market being evaluated: product category, customer segment, geography, and time horizon (3-year / 5-year). Reject compound definitions; if the team is conflating two different markets, split them and analyze separately.
+
+A2. MAP THE VALUE CHAIN from raw inputs to end consumer. Identify where margin pools concentrate (use gross-margin-by-layer logic, not revenue). The target layer(s) must be explicit. If margin is concentrated in a layer the entrant cannot reach with its current assets or capabilities, flag this immediately — it is a structural disqualifier until resolved.
+
+A3. PLATFORM VS PRODUCT MARKET CHECK (mandatory before any sizing): Determine whether this market has multi-sided network effects that create winner-take-most dynamics. Indicators: the leading player's margin expands as share grows (platform leverage), switching costs compound with usage (lock-in), or the second-place player earns structurally less margin than the leader despite similar revenue. If platform dynamics are present, the entry calculus shifts to: "Can we win the platform, or must we build an adjacent one?" Analyze accordingly. If absent, proceed as a product market.
+
+A4. TIMING WINDOW — WHY NOW: Identify the structural event, regulatory shift, technology discontinuity, or incumbent misstep that makes this window different from 18 months ago and 18 months from now. Valid timing arguments include: incumbent cost structure disrupted by input cost shift, regulatory change creating a new product category, a dominant player distracted by internal restructuring, or a distribution channel opening (e.g., a new app store, a new procurement framework). "We are ready now" is not a timing argument — it describes the entrant, not the market.
+
+A5. Distinguish the Serviceable Addressable Market (SAM) from the Total Addressable Market (TAM): TAM is theoretical ceiling; SAM is what the entrant's business model can realistically reach given channel, price point, and regulatory access. Use SAM for all sizing decisions; flag TAM only to contextualize headroom.
+
+A6. Confirm market boundaries by identifying the 3-5 adjacent markets that buyers could substitute toward. If substitution cost is low, the effective market is larger but pricing power is lower — resolve this tension explicitly.
+
+A7. Document primary sources used (regulatory filings, industry association data, disclosed financials of public comps, customer interviews). Note any data gaps and their materiality to the analysis.
+
+═══ PHASE B — MARKET ATTRACTIVENESS ═══
+
+B1. SIZE & GROWTH: State the current SAM in dollars/units with source and year. State the CAGR (historical 3-yr and forward-consensus). Flag if growth is driven by price inflation vs unit volume expansion — these have different strategic implications. A market growing only via price inflation is far more vulnerable to disruption than one growing via unit expansion.
+
+B2. STRUCTURAL ANALYSIS (Porter's Five Forces, applied rigorously):
+   - Buyer power: concentration, switching cost, price sensitivity, private-label threat.
+   - Supplier power: concentration, substitutability, forward-integration risk.
+   - Competitive rivalry: number of credible players, price vs non-price competition, capacity utilization trends, exit barriers. Note whether rivalry is stable or currently intensifying — intensifying rivalry destroys the returns that made the market look attractive from the outside.
+   - New entrant threat: capital intensity, regulatory moat, network effects, minimum efficient scale.
+   - Substitution: technology displacement horizon, cross-industry threats.
+   Score each force (Low / Medium / High) and derive an overall structural attractiveness verdict. If three or more forces score High, the market is structurally unattractive regardless of its size; document this conclusion explicitly.
+
+B3. REGULATORY & MACRO ENVIRONMENT: Identify the 2-3 regulatory levers most likely to shift the competitive landscape (tariffs, licensing, data-localization, ESG mandates, antitrust scrutiny of incumbents). Rate as Tailwind / Headwind / Neutral with a 3-year forward view. For each Headwind, estimate the capital or time cost of compliance and whether it affects the entrant asymmetrically vs incumbents.
+
+B4. CYCLE POSITION: Determine where the market sits in its adoption S-curve (emerging / growth / maturity / decline).
+   - Emerging: category-creation strategy; win by defining the buying criteria before incumbents.
+   - Growth: land-and-expand with a differentiated wedge; price competition is secondary.
+   - Maturity: displacement strategy only — the entrant must take share, not grow with the tide. A displacement entry requires a cost advantage of at least 20-30% vs the incumbent's marginal cost, or a differentiation axis the incumbent structurally cannot match (e.g., regulatory conflict, channel conflict with their existing business, or a technology generation change).
+   - Decline: acqui-hire or asset-strip logic only; organic entry is almost never justified.
+   Misreading cycle position is one of the most common and expensive entry errors — cross-check with unit volume growth (not revenue growth) and gross margin trends of incumbents.
+
+B5. PROFITABILITY BENCHMARK: Pull disclosed EBIT margins of 3+ established players in this market. If industry median EBIT < 10%, document the structural cause (commodity input pricing, buyer concentration, regulatory price caps, etc.) before proceeding. Thin-margin markets require a fundamentally different cost architecture to justify entry — "we'll be more efficient" is not sufficient; the specific cost advantage mechanism must be named.
+
+═══ PHASE C — RIGHT-TO-WIN ═══
+
+C1. CAPABILITY AUDIT: List the 5-7 capabilities that determine competitive success in this market (e.g., manufacturing tolerance at ±5 microns, regulatory relationship depth with FDA or FCA, last-mile logistics density in Tier-2 cities, proprietary behavioral dataset with >10M labeled events, brand trust measured by NPS differential vs category average). For each, rate the entrant's current capability vs the leading incumbent: Ahead / Parity / Lagging / Absent. Capabilities rated Absent that are listed as critical success factors are entry-blocking unless a specific acquisition or partnership plan addresses them.
+
+C2. ASSET INVENTORY: Identify proprietary assets (IP, exclusive relationships, datasets, geographic footprint, brand equity) that are both material to winning AND not easily replicated in <3 years. For each asset, state: the specific advantage it confers, the mechanism that makes it durable (legal protection, data network effect, switching cost, scale-dependent unit economics), and the estimated time for a well-funded competitor to replicate it. Assets replicable in <18 months are temporary advantages — weight them at 50% of face value in the right-to-win assessment.
+
+C3. COST POSITION AND STRATEGIC DIFFERENTIATION PATH: Model the entrant's unit economics at 3 scale points (launch, Year-2 run rate, steady state). Compare explicitly to estimated incumbent cost structure. Three and only three sustainable paths exist:
+   (a) Structural cost advantage: the entrant's architecture, input mix, or operating model achieves lower marginal cost at steady state — name the specific mechanism (e.g., zero-field-sales distribution, proprietary manufacturing process, software-defined delivery).
+   (b) Willingness-to-pay premium: customers will pay a price premium because the entrant delivers measurably superior outcomes on an axis the incumbent cannot match without cannibalizing its existing business (e.g., compliance-native product in a regulated vertical, integration value that reduces customer total cost of ownership even at a higher unit price).
+   (c) Niche cost-to-serve advantage: the entrant serves a segment the incumbent over-serves (paying for capabilities the niche does not value), allowing a stripped-down, lower-cost offering that is still superior on the dimensions that matter to that niche.
+   If none of these three paths is credible, entry is not viable at current capabilities.
+
+C4. CUSTOMER ACQUISITION PATHWAY: Identify the first 10 target customers or customer archetypes by name or precise description. For each, name: the decision-maker role and their primary KPI, the incumbent they currently use, the specific switching trigger (what event or pain threshold would cause them to evaluate an alternative), and the realistic sales cycle duration. Generic demand assumptions derived from TAM are disallowed here. If 10 specific targets cannot be named, the beachhead is not yet defined.
+
+C5. COMPETITIVE RESPONSE MODELING: For the top 2 incumbents, model their most likely defensive moves within 12 months of the entrant's market announcement: price war, product bundling, exclusivity deals with shared channel partners, counter-acquisition of the entrant's suppliers or distributors, or regulatory lobbying. For each defensive move, assess the entrant's resilience. If the entrant cannot survive a 12-month price war from an incumbent with 3x margins, that is a structural vulnerability that must be disclosed — not buried in an appendix.
+
+C6. RIGHT-TO-WIN VERDICT: Issue a binary verdict (Credible / Not Credible) with a one-paragraph rationale that directly ties back to C1-C5. A "Credible with conditions" verdict must name the conditions as specific, time-bound actions (e.g., "Credible only if exclusive distribution agreement with [partner type] is signed before $X of capital is deployed, by [date]").
+
+═══ PHASE D — ENTRY MODE, SEQUENCING, AND COMMITMENT CLASSIFICATION ═══
+
+D1. ENTRY MODE OPTIONS: Evaluate at minimum these four modes for this specific market and entrant profile:
+   (a) Organic build — own product/team/channel from scratch.
+   (b) Acquisition or acqui-hire — buy capability, customer base, or license.
+   (c) Partnership or JV — share risk, leverage incumbent distribution or regulatory standing.
+   (d) White-label / OEM / licensing — asset-light, margin-thin, speed-maximizing.
+   For each mode, estimate: time-to-first-revenue, capital required to reach breakeven, degree of strategic control retained, and exit optionality if the entry fails. The preferred mode should score highest on the axis that matches the entrant's binding constraint (time, capital, or regulatory access).
+
+D2. ENTRY SEQUENCING — BEACHHEAD DEFINITION: Define the beachhead as the narrowest defensible segment where the entrant can achieve market leadership (>20% share) within 24 months before expanding. The beachhead must be chosen to maximize learning rate and minimize incumbent retaliation — not to maximize near-term revenue. A beachhead that immediately threatens an incumbent's core business will trigger a defensive response that few entrants can absorb. The ideal beachhead is a segment the incumbent is structurally unable or unwilling to defend.
+
+D3. COMMITMENT CLASSIFICATION — for every major commitment in the entry plan, classify explicitly:
+   IRREVERSIBLE (>18 months to unwind, capital above the board-approved threshold, or reputational lock-in that affects the parent entity): treat as a strategic bet requiring a formal pre-mortem and board-level documentation.
+   REVERSIBLE (can be exited within 6 months at <20% sunk cost): treat as a funded experiment with explicit success metrics set before deployment.
+   List at least 3 of each type. If the entry plan has more than 2 irreversible commitments in Stage 1, the staging is wrong — restructure to push irreversible commitments to later stages.
+
+D4. STAGED CAPITAL PLAN: Define 3 funding gates (Stage 1 / Stage 2 / Stage 3). For each gate, state: the capital deployed, the specific hypothesis being tested, the metric that confirms the hypothesis at a specific threshold, and the go/no-go decision rule. No gate should be defined as "when we feel ready" or "when the market responds." Stage 1 capital should be sized to answer the most important unknown identified in Phase F (learning agenda), not to achieve revenue targets.
+
+D5. OPERATING MODEL AT ENTRY: Name the specific legal vehicle (e.g., branch office vs local subsidiary, regulated entity vs holding company structure), the country/state of incorporation and its tax/liability implications for this specific market, the initial headcount shape (not just a number — which functions are in-house vs contracted, and why), and the first 3 roles that are non-negotiable for execution. For market entry specifically, the non-negotiable roles are almost always: the person who owns regulatory relationships or channel access, the person who owns the first customer relationships, and the person who owns the operational model that must work at 10x launch volume. If these three archetypes are not covered by the first 3 hires, the operating model is wrong.
+
+═══ PHASE E — STRONGEST REASONS NOT TO ENTER ═══
+
+E1. Identify and develop the 2-3 strongest arguments AGAINST entry. These must be structural arguments — meaning they would persist regardless of how well the team executes and regardless of how much capital is deployed. Examples of structural arguments: the market's margin pool concentrates in a layer the entrant's business model cannot access; the dominant player's data or network effect compounds faster than any new entrant can acquire customers; the regulatory approval timeline structurally exceeds the window of market opportunity; the incumbent's channel relationships create a distribution moat that cannot be replicated without the very scale the entrant lacks (a catch-22). Execution risks (we might hire badly, we might be slow) do not qualify as structural arguments.
+
+E2. For each structural argument, assess: (a) Is it a permanent barrier or a time-limited window that will close regardless of what the entrant does? (b) Can it be mitigated by entry-mode selection or beachhead sequencing, or is it mode-agnostic and beachhead-agnostic? (c) What specific evidence — observable in advance — would cause you to revise this argument?
+
+E3. DEVIL'S ADVOCATE SCENARIO: Describe the most plausible scenario in which the entrant enters, executes competently, and still fails. Name the specific mechanism of failure (e.g., the incumbent responds by bundling the entrant's product category into its platform at zero incremental price, eliminating the entrant's ability to charge; the regulatory approval arrives 18 months late, after the market has consolidated around a competitor that moved earlier). If this scenario has a probability above 30%, the entry thesis requires a fundamental restructure before proceeding — not cosmetic risk mitigations.
+
+═══ PHASE F — STAGED DE-RISKING PLAN ═══
+
+F1. KILL CRITERIA (pre-register before any capital is deployed):
+   Stage 1 kill: [Metric] does not reach [Threshold] by [Date] → exit and recover remaining capital.
+   Stage 2 kill: [Metric] does not reach [Threshold] by [Date] → exit or pivot to [specific named alternative].
+   Stage 3 kill: [Metric] does not reach [Threshold] by [Date] → restructure or divest at best available terms.
+   Criteria must be set before observing results; post-hoc criteria (adjusted after early data) are disallowed and must be flagged if they occur.
+
+F2. EARLY SIGNAL INDICATORS: Identify 3-5 leading indicators measurable within 90 days of launch that predict whether Stage 1 kill criteria will be met. Leading indicators must precede the lagging metric by 60-90 days so the team has time to react before the kill trigger is reached. Examples: pilot conversion rate from evaluation to paid, inbound inquiry volume from the target segment before public launch, time-to-close in the first 5 sales cycles vs model assumption, or a supplier's willingness to offer preferential terms (a signal of perceived entrant viability).
+
+F3. PIVOT OPTIONS: For each Stage kill trigger, name the specific pivot — the alternative that the team will pursue if the trigger fires. Pivots must be scoped and pre-analyzed, not open-ended. "Find another approach" is not a pivot. Valid pivot examples: shift from direct sales to channel partnership, shift from end-customer to OEM supply, narrow beachhead to a sub-segment with demonstrated pull, or exit the market and redeploy capital to an adjacent opportunity already identified in Phase A.
+
+F4. LEARNING AGENDA: List the 5 most important unknowns that Stage 1 must resolve — the questions that, if answered wrongly, would cause the entry thesis to fail. Every dollar in Stage 1 is a payment for information, not a bet on certainty. If Stage 1 cannot resolve these unknowns because the experiment design is wrong (wrong segment, wrong product, wrong channel), fix the experiment before deploying capital. Examples of valid unknowns: Does the target customer segment actually experience this pain acutely enough to switch from the incumbent? Is our cost-to-serve model achievable at 20-unit scale, or does it require 200-unit scale to become viable? Will the regulatory body accept our compliance approach, or will it require a redesign?
+
+F5. DECISION AUTHORITY MAP: Name who has authority to call a kill at each stage and what quorum is required. If the kill decision requires consensus across more than 3 parties, the governance is too slow for a market entry context — flag this as an organizational risk with a specific proposal to streamline it. The person who has authority to call a kill must be named before Stage 1 capital is deployed; discovering authority ambiguity mid-entry is a governance failure.
+
+═══ PHASE G — SYNTHESIS AND RECOMMENDATION ═══
+
+G1. INTEGRATION CHECK — before issuing a verdict, answer these four questions explicitly, citing the phase that generated each answer:
+   (a) Is the market structurally attractive and not already past the window? (Phase A timing + Phase B)
+   (b) Does the entrant have a credible right-to-win, or does one specific condition need to be true first? (Phase C)
+   (c) Is the entry mode and beachhead matched to the entrant's binding constraint (time, capital, or access)? (Phase D)
+   (d) Have the strongest structural arguments against entry been addressed — not dismissed? (Phase E)
+   If any of the four answers is "no" or "not yet," the recommendation in G2 must be conditional, not unconditional.
+
+G2. ENTRY VERDICT: State a single, clear recommendation — Enter / Do Not Enter / Enter Only If [specific condition is met by specific date]. The verdict must follow from G1 and must not contradict any finding from Phases A-F without explicitly noting the override and its rationale.
+
+G3. RECOMMENDED MODE AND BEACHHEAD: One paragraph naming the specific entry mode (legal vehicle, partner or acquisition target if applicable), the beachhead segment defined with the specificity required in D2, and the first 90-day action sequence as a numbered list. The first 90-day sequence must include at minimum: the first irreversible commitment and its trigger date, the first leading indicator checkpoint and who reviews it, and the first customer name or archetype to be approached.
+
+G4. THE ONE IRREVERSIBLE BET: Identify the single commitment that, once made, most determines the outcome of the entry attempt. Ensure the output documents: what the bet is, what it costs to exit if wrong, what evidence would confirm within 6 months that it was the right bet, and that the decision-maker named in F5 has explicitly signed off on it.
+
+G5. MONITORING CADENCE AND ESCALATION TRIGGERS: Recommend a review frequency tied to the pace of the leading indicators identified in F2 — not to the entrant's internal calendar. Fast-moving consumer markets with 30-day sales cycles need monthly reviews. Capital-intensive B2B markets with 12-month sales cycles can tolerate quarterly reviews. Specify the escalation trigger: if a leading indicator moves more than [X%] below model in any review period, the kill authority named in F5 must convene within [Y days], not at the next scheduled review.
+
+KEY PRINCIPLE: Attractiveness tells you whether a market is worth winning; right-to-win tells you whether you can win it; timing tells you whether the window is open now or closing — a complete entry thesis requires all three, and a single fatal flaw in any one is sufficient grounds to decline or defer.