Popper Market Challenger

Falsification-oriented market thesis challenger for A-shares and U.S. markets. Use when the user asks to attack an existing market thesis, morning brief, clo...

Security Scan

The skill's instructions, requirements, and scope align with its stated purpose of decomposing and stress‑testing market theses; it is an instruction‑only skill that asks for no credentials or installs.

Purpose & Capability

The name/description match the SKILL.md: it exists to decompose a market thesis into falsifiable claims and audit them using news, data, and price action. It requests no unrelated binaries, env vars, or config paths, which is appropriate for a pure analytic instruction set.

Instruction Scope

The SKILL.md stays within the declared purpose (rewrite thesis, create 3–5 claims, audit News/Data/Price Action, give decision). However, it is open‑ended about how the agent obtains 'news', 'data', and 'price action' (web browsing, internal data connectors, or user‑provided files), which grants runtime discretion about external data sources. The instructions do not tell the agent to read unrelated files or secrets.

Install Mechanism

No install spec and no code files are present (instruction‑only), so nothing is written to disk and no third‑party packages are pulled in.

Credentials

The skill declares no environment variables or credentials (proportional to an instruction‑only analytic tool). Be aware that practical use may require market data access (price feeds, news APIs) via the agent's existing connectors — those would be external to the skill and require separate credentials.

Persistence & Privilege

always is false, user‑invocable is allowed, and disable-model-invocation is false (normal). The skill does not request persistent system presence or modify other skills or system configs.

Assessment

This is an instruction‑only skill that appears internally consistent. Before enabling it: provide a clear base thesis as input (it does not generate the base brief), avoid pasting secrets or API keys into prompts, and confirm what data connectors the agent will use to fetch 'news/data/price action' (browser, paid data feeds, or your internal docs). If you rely on external market data, only grant the minimum necessary access and validate the outputs before making trading decisions. If unsure, run it in short mode or with the agent's network/access restricted and supply the required market data manually for the first few runs.

This skill does not generate the base market brief. Its job is to test whether the existing thesis survives contact with evidence.

Use it only after a base case already exists, whether from:

• a Morning Brief
• a Close Review
• a Weekly Reset
• a handwritten thesis note
• a verbal market call from the user

Core purpose

The goal is not random contrarianism. The goal is to take a market thesis, break it into falsifiable claims, and check which parts still hold.

This skill should:

• extract the thesis actually being tested
• decompose it into 3-5 falsifiable sub-claims
• audit each claim using news, data, and price action
• identify the fastest disconfirming signals
• decide whether the base case survives, weakens, or breaks

This skill should not:

• write a second full morning brief unless the user explicitly asks
• manufacture a fake anti-consensus view just to sound smart
• act like a philosophy seminar
• replace the PM's final judgment

Operating mode

Treat the input thesis as a testable object. You are not trying to sound bearish or clever. You are trying to answer:

• what exactly is being claimed
• what evidence would support it
• what evidence would break it fastest
• whether the thesis still deserves current sizing and timing

Always assume:

• the base case may be partly right
• some claims may survive while others fail
• the most useful challenge is the one that changes sizing, confidence, timing, or narrative hierarchy

Required workflow

1. Thesis under test

Start by rewriting the base case into one concrete testable sentence.

Good examples:

• "Oil retreat plus rate stability is enough to support a broad risk-on repair in U.S. equities."
• "A-share lithium leadership is the start of a wider growth re-risking move, not just a local rebound."

Bad examples:

• "Markets are complicated."
• "There are risks on both sides."

2. Claim decomposition

Break the thesis into 3-5 falsifiable sub-claims. Each claim must be concrete enough to be tested against observable evidence.

Typical market claim categories:

• event / catalyst claim
• rates / FX / liquidity claim
• credit confirmation claim
• breadth / tape claim
• proxy basket / sector confirmation claim

Examples:

• breadth must confirm the index move
• small-caps must confirm risk-on
• credit must confirm the repair
• oil must stop rising for the inflation scare to fade
• homebuilders / retail / transports must confirm the growth story

3. Evidence audit

For each claim, check evidence in three buckets:

• News
• Data
• Price Action

Do not leave claims ungrounded. If one bucket is unavailable, say so. If all three are absent, mark the claim as effectively untested.

4. Survival status

Assign each claim one status only:

• Supported
• Weakly supported
• Untested
• Under active contradiction

Use Under active contradiction when observable evidence is already moving against the claim. Do not overuse it.

5. Falsification map

For each claim, state the fastest near-term signal that would break it. These must be concrete and monitorable.

Good examples:

• "This claim breaks if IWM / SPY rolls over while SPY holds up."
• "This claim breaks if HYG and KRE fail to confirm within 1-2 sessions."
• "This claim breaks if oil re-accelerates and 2Y yields rise together."

Bad examples:

• "This claim breaks if sentiment worsens."
• "This claim breaks if markets don't like it."

6. Strongest serious objection

After the claim audit, state the single strongest objection. This objection must point to the most fragile claim or broken transmission link. One paragraph only.

7. Historical analog filter

Use at most 1-2 analogs. Only include analogs if they sharpen falsification, timing, or risk management.

For each analog, state:

• what is similar
• what is different
• why the difference matters

Do not force analogs into every run.

8. Decision impact

End with a PM-facing conclusion. Answer:

• does the base case survive?
• should sizing stay, shrink, wait for confirmation, or be re-ranked?

Use one of these only:

• Base case survives
• Base case weakened
• Base case materially challenged

Then add one sentence on sizing / timing impact.

Default output structure

Keep the default output compressed and decision-relevant.

1. Thesis Under Test

One sentence.

2. Claim Audit

For each claim, use this mini-format:

• Claim: ...
• News: ...
• Data: ...
• Price Action: ...
• Status: Supported / Weakly supported / Untested / Under active contradiction
• Fastest falsifier: ...

Use 3-5 claims total.

3. Strongest Serious Objection

One paragraph.

4. Historical Analog Filter

1-2 analogs max.

5. Final Decision Impact

• Verdict: ...
• Sizing implication: ...
• What must be re-checked next: ...

Short mode vs deep mode

Short mode (default)

Use in normal daily workflows. Compress each claim to 1-2 lines of evidence.

Deep mode

Use only when:

• the user explicitly asks for depth
• the thesis drives a large position or aggressive add
• the market is at a regime transition
• the evidence is highly conflicted

In deep mode, make the evidence audit more explicit, but do not turn it into a second morning brief.

Signal quality rule

Prefer high-value falsification targets:

• stale event logic
• missing or broken transmission links
• breadth not confirming
• credit not confirming
• rates / FX contradicting the thesis
• proxy basket divergence
• sector leadership too narrow
• crowding / pain-trade risk

Do not waste time on low-value objections that would not change behavior.

Style

Voice should be direct, skeptical, and useful. Not theatrical. Not snarky. Not academic for its own sake.

Good tone:

• "The thesis under test is..."
• "This claim is only weakly supported because..."
• "The fastest falsifier is..."
• "The main transmission break is..."
• "The base case survives, but only if..."

Bad tone:

• vague contrarianism
• performative complexity
• philosophy cosplay
• generic "there are risks on both sides"

Final rule

A good Popper-style challenge is one that could realistically make the PM:

• cut size
• delay entry
• demand one more confirmation
• re-rank the narrative hierarchy
• or keep the thesis but narrow the expression

If the challenge would not change behavior, it is probably too weak.