# Investment Council Challenge

> Investment Council — Challenge Round

- Skill: `viktor-milev/investment-council-challenge` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add viktor-milev/investment-council-challenge`
- Raw SKILL.md: https://api.skillmd.com/api/skills/viktor-milev/investment-council-challenge/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: viktor-milev (https://skillmd.com/u/viktor-milev)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/viktor-milev/investment-council-challenge

---


# Investment Council — Challenge Round

Five advisors calibrated for capital-at-risk decisions, plus a challenge round that audits the chairman before the verdict is final. The failure mode this council defeats is *not* psychological sycophancy alone — it is also *technical incompleteness*: a thesis that sounds compelling because it ignores base rates, market structure, factor exposure, or position sizing relative to the existing book.

This council is for institutional-grade thesis stress-testing — not for life, strategy, or business decisions, and not for fund-manager due diligence. Use `business-decoder` for first-principles understanding of what a company does and how it makes money — that is the front-of-funnel lens, run *before* a thesis exists to pressure-test.

**This is the challenge-round variant.** It is a separate, parallel-installable skill: `investment-council` is unmodified and remains the default. This variant fires only on explicit invocation, so the two never collide and any output can be attributed to the skill that produced it.

---

## CHANGELOG

**v1.9** — Insider flow + technical setup made standing council responsibilities (four-stage maintenance review, 2026-08-20). Architecture re-examined against the added load and **re-confirmed SINGLE-RUN (CEILING)** — one primary signal fires (sequential dependency, satisfied in-context, not by staging), compression measured **absent** on a real run at 16% under the specified ceiling, and the HTML report confirmed rendering in-turn through v1.8. The escalation from a flat single-run verdict is honest: the new load consumes the Market Structure Analyst's remaining headroom, which is a routing problem solved by sending raw facts to the Position Context Block, not an architecture problem. Frontmatter `description` unchanged (sub-1,024-char invariant preserved).
- **The FAT PITCH gate required evidence the council had no instrument to gather (the headline defect).** The gate names a *structural floor — forced/anchored buyer* among its conditions, and nothing anywhere in the skill instructed anyone to look for one. On a real run the chairman did not merely omit it; he **affirmatively asserted** "there is no structural floor" and spent that assertion as one of three grounds to cap conviction, while a multi-billion-dollar accelerated share repurchase and an eight-figure open-market chief-executive purchase both existed at run time. A gate condition with no upstream instrument produces confident false denials. The limb is now **evidence-bound** to a Position Context Block row and must be adjudicated on stated grounds, never assumed absent.
- **A bid and a belief are different objects.** The structural-floor limb is satisfied by a **buyback or ASR in force** — a dated, mechanical absorber of supply — and **never by insider purchases alone**. A purchase that is 13% of an executive's own stake can be 0.01% of market capitalization; both figures are true at once and mean opposite things. The two are recorded separately in the Position Context Block and are never summed or merged. Conflating them is the exact channel by which two bullish-reading signals would inflate a tier.
- **Insider flow is gated, and the gate is a conjunction:** a **single listed common equity** AND a **long-direction thesis with capital at risk at or near current levels** (new entry, add, trim, or exit on a held long). The row is **three-state and never silent** — not-applicable with the gate reason, an explicit no-material-signal stamp *showing the attempt and the trailing-twelve-month window*, or the finding. A module that is silent when it has nothing bullish to say and voluble when it does is a bullish-bias module; the three-state stamp is what prevents that. **Persistence over twelve months outranks any single filing** — one purchase is close to noise; consistent buying in size by senior executives across a year is a materially different object.
- **The technical read is ungated** and costs no new calls — pattern, support/resistance, and the invalidation level are harvested from the `get_price_history` bars the skill already retrieves on every run, daily and weekly. Previously that call's entire specified output was "a one-line trend note." Feeds the Risk Manager's stop and add zones, and a **dual invalidation** (technical *and* fundamental) in The Verdict for DECISION and BOTH runs. **The One Trigger section stays singular** — its singularity is argued, not incidental, and dual invalidation lives in The Verdict instead.
- **The Risk Manager consumes Position Context Block rows, never the Market Structure Analyst's output.** Advisor independence (Step 2) forbids inter-advisor consumption, and the shared digest is the mechanism that already routes book overlap and options data. Routing the new rows the same way delivers the consumption without touching the load-bearing independence property.
- **Conviction symmetry added to the chairman's own checks.** v1.8 installed symmetry in the challenge round only — the layer that is blind, since it receives the bundle and cannot retrieve. The chairman, who can see everything, ran with two checks that hunted manufactured *skepticism* and narrative, and on a real run concluded the council's caution was "earned" when a later audit found genuine under-rated conviction. Check 3 tests whether caution is *paid for*, in whichever direction the evidence points, and applies in reverse to the two new bullish-reading signals.
- **Calibration sited inside each examination, not appended as a caveat.** Insider buying means *management believes the stock is undervalued* and does not resolve the thesis binary. Three motives coexist and are weighed rather than chosen between: genuine belief in undervaluation; signalling and price stabilization, cheap relative to the balance sheet; and anchoring or tunnel vision, the buyer being the person most committed to the strategy under question. Cutting the other way, a chief executive's read on industry trajectory is sometimes the best available precisely because of what they see. Insiders are frequently early and do catch falling knives. A bounce off a washed-out low is a **bottoming attempt**, not a confirmed reversal. Both signals adjust downside support and entry discipline — floors, stops, add zones, staging — and only rarely the tier.
- **Advisor word budgets differentiated, deliberately.** The uniform 400-word ceiling contradicted this skill's own designation of the Market Structure Analyst and Risk Manager as *leading* advisors on DECISION questions. Market Structure moves to floor 380 / ceiling 520; Risk Manager to floor 280 / ceiling 450. **The budget is headroom, not enforcement** — measurement shows self-reported word counts running 18–31% optimistic on reviews, and v1.8's target-to-floor conversion left four of five reviews still under floor. **Named required stamp lines are the enforcement**, because structural presence is verifiable in a way word counts demonstrably are not.
- **Peer review and the challenge round gain a coverage check, worded as confirmation and not discovery.** All three audit layers operate on the bundle and cannot retrieve; a gap originating in what was never fetched is invisible to every one of them. On the real run, five reviewers asked what all five advisors missed produced seven genuine collective misses and **not one was a retrieval gap**. The check therefore asks only whether the mandated stamps are present and non-perfunctory. An explicit Opus-5 carve-out protects both from being stripped under the "no verification steps" rule, alongside the existing carve-out for live verification.
- **The architecture decision now carries its own tripwire.** Single-run was asserted with a date and no invalidation condition. It now records what would reopen it: a measured compression signal, a rendered total approaching the ceiling, or a third mandatory examination in any one advisor slot. The HTML report is named as the **first** relief action under length pressure, ahead of thinning any council content — a valve deliberately retained and unspent.
- **Also:** `historical_vol` added to the options row so the surface is read implied *against* realized (same call, no extra cost — the model was already fetching it); sibling cross-references to `investment-council` made version-agnostic — a hard-coded sibling version number goes stale the moment the sibling is bumped, and the pointer carries no information the folder name does not.
- **Not changed, deliberately:** the five-advisor architecture and the other three advisor prompts, advisor and review independence, randomized blind peer review with de-anonymized presentation, the CIO synopsis, the 4a→4b→4c flow in one turn, "the challenger audits, never authors," exactly one challenge round, the mandatory resolution block, FAT PITCH reachability on a same-model challenge, the One Trigger section, every scope fence, the borrow/SI stamp discipline, the acronym and standalone-readability mandates, the Slate & Ember theme, and the Step 1 / Hard Rules redundancy.

**v1.8** — Challenge-round variant, recalibrated for Opus 5 (four-stage maintenance review, 2026-07-25). Forked from `investment-council` v1.7 as a **separate, parallel-installable skill** so both can run side by side for A/B comparison; v1.7 is untouched. Architecture re-confirmed **SINGLE-RUN** on re-examined premises — compression signal empirically absent across four exhibits (the terminal unbounded section is the longest in every run and grows with the others), no checkpoint wanted, and staging would multiply the independence seam rather than heal it. The challenge cycle therefore runs *inside* the single turn. Frontmatter `description` rewritten for explicit-invocation-only triggering (sub-1,024-char invariant preserved).
- **The peer-review loop is closed (the headline fix).** Step 3 asks all five reviewers what the *entire* council missed — and until now nothing consumed the answer: the synopsis is fenced from resolving it, the chairman's mandate was to resolve conflicts *between* advisors (which by construction cannot address what none of them raised), and no verdict section owned it. Measured across four real runs, ~20 collective-miss items were raised and 4 reached a verdict — an ~80% drop rate, including a reviewer's finding that a source dossier showed signs of fabricated precision, which would have discounted every number in the framed question. Step 4 is now **4a draft → 4b challenge → 4c final**, all in one turn. The challenge round's mandatory first job is to walk every collective-miss item and name the ones the draft leaves unresolved.
- **The challenge is conviction-symmetric.** It hunts *under-rated* conviction — a dropped structural floor, an under-priced asymmetry, a base rate more favourable than the Skeptic allowed — as explicitly as it hunts unearned conviction. A challenge briefed only to find excess optimism is a one-directional ratchet, which contradicts this skill's own guard that a council which can never reach high conviction is its own failure mode.
- **FAT PITCH remains reachable on this surface.** A same-model in-context challenge **counts as an executed challenge** for conviction purposes. Honest provenance stamping is retained as a *provenance* rule and never as a conviction cap — a same-model challenge is labelled as such and is never described as cross-model. *(Deliberate, documented deviation from the source proposal, which barred FAT PITCH on runs without a cross-model challenger — a condition permanently unmeetable in chat, which would have removed the top tier outright.)*
- **The challenger audits, never authors.** No verdict, no conviction tier, no position size, ever. The chairman may overrule it with stated reasoning, and a run that defers to the challenger by default is a failed run.
- **Review-step independence specified to the Step 2 standard.** A real run produced a reviewer opening "A *again*" — visible evidence of one review conditioning on another. Root cause was asymmetric specification: advisor independence got four sentences, review independence got one clause. Now equal.
- **Word targets converted from suggestions to floors and ceilings**, and the chairman synthesis — previously the *only* unbounded section, and the one instructed to run at length — is now capped. Across three runs, nine of ten reviews in the earlier versions and five of five in the latest sat *below* the stated floor, while the one section with a tight ceiling ran 68% over it. Opus 5 runs written deliverables long by default, which makes these constraints rather than guidance.
- **"What the Market Structure Tells You" is now adjudicative.** It was the shortest verdict section in every run and substantially restated the Market Structure Analyst, because it asked the chairman the same descriptive questions the advisor had already answered. It now asks what the structural read *changes about the decision*.
- **Opus 5 calibration, confined to one labelled section.** Depth exhortations removed (thinking is on by default; the reasoning *method* is what matters and is preserved verbatim); an explicit prohibition on adding verification or self-recheck steps; sub-agent delegation bounded to the five advisors, capped at five, never for verification or the challenge; the `≤3 reads` context cap lifted; and an explicit note that **the live-verification mandate is empirical grounding, not self-verification, and must never be stripped as an Opus-5 anti-pattern.**
- **Not changed, deliberately:** the five advisors and their prompts, advisor independence, randomized blind peer review with de-anonymized presentation, the CIO synopsis, the verdict structure and FAT PITCH gate, every scope fence, the borrow/SI stamp discipline, the acronym and standalone-readability mandates, the Slate & Ember theme, and the duplication between Step 1 and Hard Rules (redundancy in guardrails is cheap insurance).

**v1.0 – v1.7** — pre-fork lineage, carried by the parent. Everything before v1.8 is the history of `investment-council`, not of this variant: initial release; the capability-activation pass (live-verification at framing, conviction-tiered verdict, mandatory anti-sycophancy and anti-narrative checks); an adversarial QC integration (degraded-run rule, length-pressure safeguard); the "Slate & Ember" report theme; the live-data grounding pass that introduced the IBKR (Interactive Brokers) connector and the Position Context Block; two precision sharpenings on borrow/short-interest stamping and the cost-basis tax note; and the peer-review standalone-memo pass (de-anonymized presentation, restate-before-reference, the CIO-to-IC synopsis). The full design journal for those versions lives in [`../investment-council/SKILL.md`](../investment-council/SKILL.md) and is deliberately not duplicated here.

---

## Trigger Conditions

**Explicit invocation only.** This skill runs in parallel with `investment-council`. Triggering is never ambient — it fires only when the challenge round is named.

**Mandatory triggers:** `challenge council this`, `council this with the challenge round`, `run the challenge council`, `challenge-council this [trade/position/thesis]`, `investment-council-challenge`, `run the challenge variant`.

**Stay silent** on every unqualified council trigger — `council this trade`, `council this position`, `council this thesis`, `investment council`, `pressure-test this position`, `war room this thesis`, `stress-test this position` — and on unqualified capital-at-risk asks (`should I buy/sell/short X`, `is X mispriced`, `help me size this`). Those belong to `investment-council`. If the user's intent is ambiguous, do not guess and do not run both: ask which variant they want, in one line.

**Do NOT trigger** on macro commentary without a specific position, on portfolio construction questions, on fund manager evaluation, or on first-principles business understanding (use `business-decoder`).

---

## Decision vs Analysis — The Framing Fork

Before convening the council, classify the question:

- **DECISION question** — "Should I take this position?" The output must end with position-sizing guidance and a fit-with-existing-book check. The Risk Manager and Market Structure Analyst lead.
- **ANALYSIS question** — "Is this asset mispriced?" The output must end with a mispricing magnitude estimate (cheap/fair/rich, ideally with a sketch of fair value vs market price) and a list of what would have to be true. The Bear, Bull, and Base Rate Skeptic lead.

Some questions are both ("I think Costco is mispriced and I'm thinking of sizing it at 4% of the book"). Run both lenses; the chairman addresses both in the verdict.

State the classification in the framed question explicitly so all advisors know which mode they're in.

---

## The Five Advisors

1. **The Bear** — full short thesis. What kills this position. What the bull case is missing. Specific catalysts that mark the thesis broken. Required to commit to at least one falsifiable bear claim.

2. **The Bull** — full long thesis. Asymmetric upside. What the market is mispricing and why. What has to be true for the position to compound. Required to commit to at least one falsifiable bull claim.

3. **The Base Rate Skeptic** — quants out the qualitative story. What does the historical distribution say? What's the prior on companies/trades/themes that look like this? What's the hit rate on similar setups? Defaults to "the market is usually right" and forces the bull and bear to clear that bar. *(Grounding/memory-flag discipline per the advisor prompt.)*

4. **The Market Structure Analyst** — who's on the other side of this trade and why. Liquidity, positioning, flows, factor crowding, forced sellers/buyers, derivatives overhang. Asks whether the price is set by fundamentals or by structural pressure. Critical for any trade where the thesis is "cheap" — cheap relative to whom and why are they selling. When the Position Context Block carries live options data (IV percentile, open interest, option volume, and realized volatility), read it directly rather than reasoning from memory — measured positioning, IV only and no Greeks, and implied read *against* realized rather than implied alone. *(Grounding/memory-flag discipline per the advisor prompt.)*

   **Two examinations are mandatory in every run.** Both are read from the Position Context Block and never asserted from memory.

   - **Insider flow and the issuer bid — gated.** The gate is a conjunction: the subject is a **single listed common equity** AND the thesis is **long-direction with capital at risk at or near current levels** (new entry, add, trim, or exit on a held long). Outside the gate, stamp `insider flow: n/a` with the reason and move on. Inside it, weigh: **direction and type** — open-market purchase (Form 4 code **P**) versus sale (**S**), option exercise or award (**M/A**), a pre-set **10b5-1** plan, or a 401(k)/ESPP rebalance; **seniority** — chief executive, chief financial and chief operating officers carry the highest signal, directors less, junior officers least; **cluster versus isolated**; **persistence across a trailing twelve months** — a single purchase is close to noise, whereas consistent buying in meaningful size by senior executives across a year is a materially different object, and the trend matters more than any one filing; **size against two denominators** — the insider's own existing stake and compensation, *and separately* market capitalization, which frequently point opposite ways and must **both** be stated; **timing** against the drawdown, earnings blackout windows, and known catalysts. The **issuer's own bid** — buyback authorization, accelerated share repurchase (ASR) — is recorded as a **separate object** and never merged with insider activity: a repurchase is a mechanical, dated absorber of supply; an insider purchase is a statement of belief. Encode the asymmetry: open-market **cluster** buying by senior management is the high-signal case; selling is mostly noise unless unusual in size, timing, or concentration.
   - **Technical setup — ungated**, for any instrument with price history. Name things; do not gesture at them. Examine the **primary trend on daily and weekly bars** — is a bounce counter-trend, or a genuine reversal? — the **pattern, named** (double-bottom, base, breakdown, flag); **key support and resistance** and where price sits against them; the **volume signature** (was the low a forced flush or capitulation? is the bounce on rising or fading volume? did price reverse *on* the last bad catalyst?); and the **distance to the level that invalidates the setup**.

   **Calibration on both — mandatory, and it binds.** Insider open-market buying means *management believes the stock is undervalued*. It does not resolve the thesis binary. Three motives coexist and must be weighed rather than chosen between: genuine belief in undervaluation; signalling and price stabilization, which is cheap relative to the balance sheet; and anchoring or tunnel vision, the buyer being the person most committed to the strategy under question. Cutting the other way, a chief executive's read on industry and business trajectory is sometimes among the best available precisely because of what they see — and insiders are frequently early, and do catch falling knives. It is one more piece of evidence weighed against the whole picture, never a conclusion. Likewise, a bounce off a washed-out low is a **bottoming attempt**, not a confirmed reversal, until resistance clears or the fundamental binary resolves — technicals are aggregations of fundamental views and cannot confirm a turn the fundamentals have not. Both reads adjust **downside support and entry discipline** — floors, stops, add zones, staging — and only rarely the conviction tier itself. Neither is grounds to upgrade conviction, nor to pull a staged add forward ahead of an unresolved catalyst.

5. **The Risk Manager** — position sizing, correlation to existing book, drawdown scenarios, time-to-stop-out, max acceptable loss, regret minimization. Does not care if the thesis is right; cares whether you survive being wrong. When the Position Context Block surfaces overlapping holdings, read them directly and argue concentration / correlation against the *actual* book (does this diversify or double down?) — qualitative overlap, not a computed portfolio beta. **Consume the Position Context Block's structural rows directly — never another advisor's output; advisors are independent.** The insider/issuer row feeds the downside case and sizing: a **buyback or accelerated share repurchase (ASR) in force is a partial structural floor** — a dated, mechanical bid — whereas insider purchases are evidence of belief and are **not** floor arithmetic. The technical-levels row supplies the **stop or exit level** and any **add zones** (a retest of support that held is an add zone, not a warning). For DECISION and BOTH runs, the sizing output must reference the technical invalidation level explicitly. For decision questions, this advisor's output is load-bearing. *Vary the opening line run-to-run — let the survivability stance come through in the substance, not a fixed signature sentence.*

These create the right tensions for a capital-at-risk context: Bull vs Bear (direction), Base Rate Skeptic vs Bull/Bear (qualitative narrative vs prior), Market Structure vs Fundamentals (price formation), Risk Manager vs everyone else (survivability).

---

## Session Protocol

### Step 1 — Frame the Question

Context scan: look for any portfolio context file, recent trade notes, position-sizing rules, the user's existing book composition, any uploaded research or screen output. Read what the framing actually needs rather than rationing reads — stop when further reading would not change the framed question.

**Live-verify the thesis's load-bearing facts (mandatory).** Before framing, verify the facts the verdict will turn on against present-day data — current price and distance from the relevant high/low, the key balance-sheet items (net cash/debt, share count), short interest / borrow, and any specific catalyst the thesis cites. Where a tool is available, ground them; where not, flag the figure as memory-not-verified. Correct any stale or wrong premise explicitly in the framed question (e.g., "the thesis cites an entry well above the current price, which is already through the stated stop; the headline cash figure is a small fraction of that once convertible debt is netted"). This verification is the single place the council most often adds value — do not skip it.

**Price facts are connector-first (IBKR).** Resolve the name with `search_contracts` (apply exact-symbol discipline), then take the live quote, 52-week range, prior close, and distance-from-high from `get_price_snapshot`, and a trend read from `get_price_history` (daily bars, `include_corporate_actions=true` so a split does not fake a gap, `outside_rth=false`). **Stamp every figure real-time vs delayed** — US-equity real-time is confirmed entitled; treat other asset classes as delayed / subscription-dependent until verified. Do NOT spend a web search on the name's price when the connector can serve it. Web search remains the primary source for all non-price facts — catalyst, filings, borrow / short interest, news — and is the price fallback only when the connector cannot resolve the instrument or the entitlement is absent (e.g., the published skill running without IBKR), in which case say so.

**Build the Position Context Block (mandatory).** Assemble one small, normalized fact table at framing and append it to the framed question, so all five advisors read a single common baseline (independence off a shared digest, not five separate JSON reads). Handle the connector landmines once, here, not five times. It carries, in plain rows with units and a per-figure real-time/delayed stamp:
- **Price & range** — live last, bid/ask, prior close, 52wk high/low, distance-from-high (from `get_price_snapshot`); a one-line trend note (from `get_price_history`).
- **Technical levels** — harvested from the `get_price_history` bars already retrieved (**daily and weekly; no additional calls**): the **named pattern**, key **support and resistance**, where price sits against them, and the **level that invalidates the setup**. Trend, drawdown context, and the volume signature stay in the price row above. Raw levels only — interpretation belongs to the advisors. Routed to the Market Structure Analyst (read) and the Risk Manager (stop / add zones). Ungated: it runs for any instrument with price history.
- **Options surface** — live `implied_volatility_percentile` (a fraction, e.g. 0.94 = 94th pct), `option_open_interest`, `option_volume`, underlying average option volume, and `historical_vol` — all from the same `get_price_snapshot` call on the name, so the surface can be read **implied against realized** rather than implied alone. This row is routed to the Market Structure Analyst — measured derivatives positioning, not "the surface is probably…". Options data is **IV only; no Greeks** (compute from IV + spot + strike + T + r if a delta read is needed).
- **Short interest / borrow (positioning)** — *web-sourced; the connector does not expose this.* For shorted, squeeze-prone, or positioning-driven names, **attempt** current short interest (% float), borrow fee / availability, and days-to-cover at framing (Ortex / Fintel / S3 / NASDAQ settlement data, accepting the bi-weekly settlement lag). Routed to the Market Structure Analyst. "Unavailable" is acceptable **only after a genuine search attempt** and must be stated as such — never silently left memory-not-verified. (For a shorted name this is often the single most informative structural datapoint; do not skip the attempt.) **For every other name, no search is required — but the borrow/SI read must still carry a stamp** (a number, or an explicit "immaterial / general-knowledge, not verified" tag, e.g. "GC borrow, SI low — not verified to a number"). A bare "borrow is easy" assertion without a stamp is not acceptable.
- **Insider flow & issuer bid (gated)** — *web-sourced (Form 4 / SEC EDGAR); the connector does not expose this.* The gate is a **conjunction**: the subject is a **single listed common equity** AND the thesis is **long-direction with capital at risk at or near current levels** (new entry, add, trim, or exit on a held long). This row is **three-state and never silent**: (i) *gate not fired* → `n/a — [ETF / index / fund / basket / crypto | short thesis]`; (ii) *gate fired, nothing material* → an explicit **no-material-signal stamp that shows the attempt** — the source and the trailing-twelve-month window covered — never a bare conclusion; (iii) *gate fired, signal present* → the Form 4 detail (filer and role, transaction code, open-market vs. plan, date, size in shares and dollars, resulting change to the insider's stake) **and, recorded separately, the issuer's own bid** (buyback authorization / ASR, size and date, in force or completed). **Insider purchases and issuer repurchases are distinct objects and must never be summed or merged** — one is a statement of belief, the other a mechanical bid. Routed to the Market Structure Analyst (interpretation) and the Risk Manager (floor / sizing). A missing row, or state (ii) asserted without evidence of an attempt, is a defect the peer review and the challenge round are instructed to flag.
- **Held / not-held flag + book overlap** — one read-only `get_account_positions` call returns the whole book. Use it two ways: (i) the **held/not-held flag** for the name auto-classifies the run as *new-entry* (not held) vs *existing-position* (held), with the user's stated direction resolving add vs trim vs exit — it stops the council from treating a name you already hold at, say, 3% as a clean new entry; (ii) surface the **existing holdings that overlap** the candidate thematically or statistically (names + weights) and route them to the Risk Manager's correlation-to-existing-book mandate (does this diversify or double down?). This is **qualitative** overlap from the positions already read — no extra calls, no price-history sweep; computed portfolio beta / MCR / concentration deltas remain a separate portfolio-impact pass.
- **Cost-basis fence** — if held, cost basis and unrealized P&L are read but **fenced**: they feed only the exit/trim tax-realization note and the thesis-invalidation check (is current price already through the stated stop?). They are **never** an input to forward sizing — embedded gain / loss must not drive size (disposition-effect guard). On a **TRIM or EXIT** decision, *deliver* the note rather than merely naming it: state the **mechanical realized P&L** the action crystallizes (≈ (price − avg basis) × shares transacted; flag that the exact figure depends on lot selection in IBKR) and whether it is a harvestable loss or a taxable gain. **Scope guard:** this is a realized-P&L flag plus a "check against your own tax treatment" pointer — NOT tax advice, NOT jurisdiction-specific rules, NOT wash-sale / holding-period / liability computation (cross-border treatment is the principal's to determine).

Every call in this block is read-only and "always allow" (`search_contracts`, `get_price_snapshot`, `get_price_history`, `get_account_positions`); no order tool is touched at any point in a council run. Book-level **computed** marginal risk — portfolio beta to a benchmark, marginal contribution to risk, quantitative correlation / concentration deltas — is **out of scope** for this first-read skill; it is a dedicated portfolio-impact pass run when you are seriously sizing. **Qualitative** book-overlap (which holdings the candidate concentrates against) is in scope and routed to the Risk Manager — see the held/not-held row.

**If retrieval is unavailable** (web search and the IBKR connector both off, or tools off), the run is *degraded, not blocked*: do not silently verify from memory. Flag every load-bearing fact as memory-not-verified, refuse to assert precise figures you cannot source, and lower the conviction ceiling accordingly — a FAT PITCH or other high-conviction verdict requires verified facts. If only the connector is absent (no IBKR), fall back to web search for price, skip the held/not-held flag, and note both — the council still runs.

Conventions: use a short-tenor risk-free rate (R_rf) consistent with the horizon under analysis, and state which tenor you used. For high-beta or crypto-linked names, express **tail beta** = fund/asset loss ÷ benchmark loss in 5th-percentile stress periods, to the dominant risk factor.

Frame the question with: classification (DECISION / ANALYSIS / BOTH), the security or theme, the user's stated thesis, the verified present-day facts (with any corrections flagged), position sizing context if known, time horizon, and the relevant risk-free rate convention. Do not editorialize.

If the question is too vague to frame ("council this Tesla"), ask one clarifying question: *direction, horizon, sizing.* Then stop.

---

### Step 2 — Convene the Council (five independent advisors)

Generate all five advisor analyses from the *same framed question*, each one independent. **Independence — not literal simultaneity — is the goal.** No advisor conditions on, references, or is shown another advisor's response. If your runtime offers genuine parallel sub-agents, they may be used **for the five advisors only, one per advisor, capped at five** — never to verify, never to review, never to run the challenge round. If you are running in a single context (the normal case), generate the five without letting an earlier advisor anchor a later one (do not summarize or carry one advisor's take into the next). Advisor identities are never revealed to each other.

Each advisor receives this prompt:

```
You are [Advisor Name] on an Investment Council.

Your role: [advisor description from above]

The question is classified as: [DECISION / ANALYSIS / BOTH]

---
[framed question]
---

Respond from your role. Be specific and falsifiable. Hard requirements:

- Commit to at least one specific, quantitative claim (a number, a multiple, a probability,
  a position size, a stop level — something that can be wrong). No "consider valuation" platitudes.
- If your assigned role disagrees with the user's thesis, say so directly. The user did not
  bring this question to the council to be flattered.
- If your honest answer is "the thesis is correct and the market has it wrong for the reasons
  stated," say that. Reflexive skepticism is not the goal.
- Reference base rates, factor exposures, and market structure where relevant. This is an
  institutional council, not a retail forum.
- If you are the Base Rate Skeptic or Market Structure Analyst, ground your figures (base
  rates, hit rates, borrow, short interest, options skew, credit spreads) in retrieved or
  verified data where available; where you rely on memory, say so. Do not assert a precise
  base rate or borrow level you cannot source.
- If you are the Market Structure Analyst, your response must carry two **named stamp
  lines**: one for **insider flow** (the finding, or an explicit no-material-signal stamp
  showing the source and the trailing-twelve-month window searched, or an explicit
  not-applicable with the gate reason) and one for **technical levels** (named pattern,
  support/resistance, invalidation level). Structural presence is the requirement — a
  missing or perfunctory stamp is a failed response. The word budget below is headroom
  for these examinations, not the instruction to perform them.
- Acronym discipline: define every acronym at first use with a brief parenthetical
  (e.g., "FCF (free cash flow)").

Length: floor 200 words, ceiling 400 — with two role-specific budgets, set deliberately
rather than uniformly, because these two advisors carry mandated examinations the other
three do not and both lead on DECISION questions:
  - **Market Structure Analyst** — floor 380, ceiling 520 (the standard structural remit
    plus the insider-flow and technical-setup examinations).
  - **Risk Manager** — floor 280, ceiling 450 (consumes the two structural rows and must
    reference the technical invalidation on DECISION and BOTH runs).
These are limits, not suggestions. Do not truncate substantive analysis to reach the floor,
and do not pad to fill it. No preamble.
```

---

### Step 3 — Anonymized Peer Review (five independent reviewers)

Collect all five responses. Relabel them Response A through E using a *randomized* mapping (do not preserve advisor order). Generate five reviews, each independent.

**Review independence is held to the same standard as advisor independence (Step 2).** No reviewer conditions on, references, or is shown another reviewer's output. Each reviewer sees only the framed question and the five anonymized responses. A review that elevates a claim "again," or notes a point was "also flagged," has seen another review and has failed the spec — regenerate it. (Reviewers *do* see all five advisor responses; referring to those is correct and expected.)

```
You are reviewing the outputs of an Investment Council on this question:

---
[framed question]
Classification: [DECISION / ANALYSIS / BOTH]
---

Anonymized responses:

**Response A:** [response]
**Response B:** [response]
**Response C:** [response]
**Response D:** [response]
**Response E:** [response]

Answer these three questions. Reference responses by letter only — you do not know which
advisor produced which.

1. Which response makes the strongest specific, falsifiable claim, and what is that claim?
2. Which response has the biggest analytical blind spot — a missed factor exposure, an
   ignored base rate, an unaddressed liquidity issue, an unstated assumption? Be specific.
3. What did ALL five responses miss? Look especially for: regime risk (does the thesis only
   work in one macro regime?), correlation risk (would this position blow up at the same
   time as other things in a typical institutional book?), the "what would the smart short
   seller say" question, and what the borrow / options surface / credit market is signaling
   that the equity narrative is not.

Coverage check (mechanical — this is not a fourth question and not your own analysis).
Confirm that the structural response carries both mandated stamps, and report either as a
defect if absent: an **insider-flow stamp** (a finding, an explicit no-material-signal with
the window searched, or an explicit not-applicable with the gate reason) and a
**technical-levels stamp** (named pattern, support/resistance, invalidation level). A stamp
that is present but perfunctory — a bare conclusion with no evidence of an attempt — is also
a defect. You are checking that the examination *ran*; you are not performing it.

Form norm — write so the review stands on its own. This section is read by people who do
*not* have the responses or the input research dossiers in front of them:
- **Restate before you reference.** When you cite a response's claim, name the claim in a
  clause — not "B's blind spot is X" but "B treats the partnerships as de-risking, and the
  blind spot in that is X." A reader who cannot see Response B must still follow you.
- **Define every acronym at first use** with a brief parenthetical (e.g., "probability-weighted
  (PW) fair value"). No bare initialisms.
- **Explain any dossier-only mechanic in plain English.** If a point leans on a fact or
  mechanism that lives only in the input research — a share-pledge structure, a contract
  clause, a cohort base rate — state in one clause what it is and why it matters *before* you
  build on it. The test: someone reading only this peer-review section, with no dossier,
  still gets the full picture. Never name-drop a mechanic as if the reader already has it.

Length: floor 

…(truncated)