agentleFS
ratacatratacat/ratacats-skills/skills/pm-market-analysis/SKILL.md

pm-market-analysis

Skill9 starsChanged 5 months agoMITView on GitHub

Analyze one PM event for PMKNB: identify exact Polymarket/Kalshi instruments (conditionId, token IDs, tickers), assess market forces, resolution/oracle risk, liquidity, and counterparty risk; output forecast, report, brief, proposal, or no-publish.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

27 rules. About 0 are common, 20 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈20

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Use PMKNB terms: situation, source, claim, instrument, forecast, position, report, brief, proposal, run_trace, and PM event.

    line 19
  2. Distinctive:

    Use instrument, not contract.

    line 21
  3. Distinctive:

    Use these as the core trader-facing lens.

    line 25
  4. Distinctive:

    Read the rules, not the title.

    line 30
  5. Distinctive:

    Never rely on title-only matching.

    line 56
  6. Distinctive:

    Do not attach Polymarket fields to Kalshi instruments.

    line 59
  7. Very specific:

    If exact identity cannot be verified, choose proposal or no_publish.

    line 61
  8. Distinctive:

    Prefer primary sources for rules, resolver/oracle details, market identity, and settlement mechanics.

    line 65
  9. Very specific:

    Avoid buy/sell/hold guidance, orders, executions, or dense orderbook history.

    line 75
  10. Very specific:

    If resolution risk dominates, prefer report, brief, proposal, or no_publish over forecast.

    line 81
  11. Distinctive:

    Use current snapshot data only if fresh.

    line 84
  12. Distinctive:

    Do not provide order placement, execution tactics, or dense orderbook history.

    line 86
  13. Distinctive:

    Do not speculate about hidden solvency, private counterparties, or non-public platform conditions.

    line 90
  14. Distinctive:

    Do event-level math only after verifying a same-venue complete MECE set.

    line 94
  15. Distinctive:

    Treat cross-venue markets as related-only unless verified rules and identity make a narrower relationship explicit.

    line 95
  16. Distinctive:

    Do not net, merge, or imply equivalence across venues.

    line 96
  17. Very specific:

    Use pm_market_analysis.v1 when structured output is requested.

    line 106
  18. Distinctive:

    Put evidence before conclusions.

    line 107
  19. Distinctive:

    When working inside a PMKNB runner mode, respect runner-owned writes: inner agents should produce the requested output or apply batch for …

    line 125
  20. Distinctive:

    When this skill is used from a PMKNB workflow runner:

    line 129
  21. Distinctive:

    Read world context first, then market state.

    line 131
  22. Distinctive:

    Verify the child market creation/listing timestamp and the timestamp of every alleged qualifying occurrence.

    line 133
  23. Very specific:

    If no exact instrument is available, emit a proposal or no_publish.

    line 134
  24. Very specific:

    If the trusted market provider is unavailable, emit a blocked result with reason market_provider_unavailable.

    line 135
  25. Very specific:

    If a readable synthesis is needed, use a report or brief product.

    line 139
  26. Distinctive:

    Every meaningful runner-backed package should include one run trace with provider status, instrument id, forecast id if written, snapshot …

    line 141
  27. Distinctive:

    When uncertain, preserve PMKNB integrity over completeness: mark uncertainty, fail the relevant gate, and choose report, proposal, or …

    line 190

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

PM Market Analysis

Purpose

Analyze one PM event for PMKNB and produce a concise, evidence-first package. A PM event may contain multiple child markets, but the analysis must center on one event identity and output exact PMKNB instruments.

Use PMKNB terms: situation, source, claim, instrument, forecast, position, report, brief, proposal, run_trace, and PM event.

Use instrument, not contract. Use PM event, not venue event. Do not use memo as a product term for this skill.

Practical Trading Heuristics

Use these as the core trader-facing lens. Keep them practical and human-readable; do not turn them into wallet, API, or execution-system details unless the user is explicitly building execution infrastructure.

  • Watch for crowded obviousness. If the reason is easy to see, assume fast traders or market makers have probably seen it too.
  • Have an exit reason before entering. Is this a hold-to-resolution trade, a catalyst trade, a mispriced-spread trade, or just a research watch? Those are different.
  • The best trades often start with source advantage. Know where the resolving fact will appear: docket, official site, clerk, agency feed, certification board, box score, or another concrete proof channel.
  • Read the rules, not the title. A market can be right in the world and still resolve differently because of source, deadline, wording, or proof standard.
  • Know what price you can actually get. Displayed price is not executable price. Spread and depth matter more than headline probability.
  • Size matters. A good-looking price may only exist for tiny size. Before treating a market as attractive, ask what the effective price is for the amount that would actually be traded.
  • Timing is part of the trade. For occurrence-based markets, compare the qualifying act's timestamp with the exact child market's createdAt/listing time. An act completed before market creation is ineligible unless the written rules explicitly apply retroactively; later reporting, confirmation, or continuing consequences do not create a new qualifying act. Also assess deadlines, dispute windows, and close times.

Load The Reference

Load references/pm-market-analysis-reference.md before producing any publishable package.

Also load it whenever the task involves:

  • Polymarket identity, conditionId, token IDs, outcomes, negative risk, or sibling markets.
  • Kalshi identity, event_ticker, market ticker, or YES/NO side.
  • Resolution/oracle ambiguity, dispute/challenge status, or rule/source mismatch.
  • Liquidity, spread, depth, VWAP, stale/pinned pricing, or event-window liquidity shock.
  • A structured pm_market_analysis.v1 output or PMKNB write matrix.

Workflow

  1. Frame the situation.
    • Identify the subject/world question.
    • Identify exactly one PM event.
    • State whether the event has one child market or multiple child markets.
    • Separate world outcome probability from resolution probability.
  2. Verify instrument identity.
    • Never rely on title-only matching.
    • For Polymarket: verify event -> market -> conditionId -> CLOB token IDs/outcomes.
    • For Kalshi: verify event_ticker -> market ticker -> YES/NO side.
    • Do not attach Polymarket fields to Kalshi instruments.
    • Record the exact child market's creation/listing timestamp, not only the parent PM event's timestamp.
    • If exact identity cannot be verified, choose proposal or no_publish.
  3. Ingest evidence before conclusions.
    • Capture sources, claims, market data snapshots, and rules.
    • Prefer primary sources for rules, resolver/oracle details, market identity, and settlement mechanics.
    • Record snapshot freshness for prices and liquidity.
    • Timestamp every alleged qualifying occurrence and compare it with market creation before using it in the thesis.
    • Mark stale, ambiguous, conflicting, or unverified claims explicitly.
  4. Analyze market forces.
    • Build an outside view.
    • Identify causal forces, clashing forces, cruxes, scenarios, falsifiers, and coherence checks.
    • Distinguish probability, confidence, and actionability.
    • Apply the practical trading heuristics: crowded obviousness, source advantage, rules-vs-title risk, executable price, sizing, timing, and exit reason.
    • Avoid buy/sell/hold guidance, orders, executions, or dense orderbook history.
  5. Assess resolution/oracle risk.
    • Inspect resolver, rules, source finality, ambiguity, challenge/dispute status, and title/rules/source mismatch.
    • Apply temporal eligibility: exclude pre-creation occurrences unless the written rules explicitly state they count retroactively. Do not reset an occurrence's time because it was reported, confirmed, implemented, or remained in effect later.
    • Separate what will happen in the world from how the instrument will resolve.
    • If resolution risk dominates, prefer report, brief, proposal, or no_publish over forecast.
  6. Assess liquidity and microstructure.
    • Use current snapshot data only if fresh.
    • Assess spread, depth/VWAP by notional, asymmetry, stale/pinned flags, and event-window liquidity shock risk.
    • Do not provide order placement, execution tactics, or dense orderbook history.
  7. Assess platform/counterparty risk.
    • Disclose settlement, custody, oracle, regulatory, and access risk.
    • Do not speculate about hidden solvency, private counterparties, or non-public platform conditions.
  8. Handle multi-market and negative risk.
    • List sibling markets separately.
    • Do event-level math only after verifying a same-venue complete MECE set.
    • Treat cross-venue markets as related-only unless verified rules and identity make a narrower relationship explicit.
    • Do not net, merge, or imply equivalence across venues.
  9. Choose the product.
    • forecast: immutable thesis revision for exact verified instruments.
    • report: evidence-led analysis without a new immutable forecast.
    • brief: short operational/context package.
    • proposal: recommendation to create/update PMKNB records, instruments, or analysis tasks.
    • no_publish: required when identity, rules, sources, or validation gates fail badly enough that the package would mislead.
  10. Emit the PMKNB package.
    • Use pm_market_analysis.v1 when structured output is requested.
    • Put evidence before conclusions.
    • Include venue-specific discriminated instrument identity.
    • Include validation gates and failed gates.
    • Include the PMKNB write matrix.
    • Include run_trace.

Product Decision Rule

Produce a forecast only when exact instrument identity, resolution rules, temporal eligibility, source basis, and minimum evidence gates pass.

Produce a report when analysis is useful but forecast conditions are not met.

Produce a brief when the user needs compact situation/source/claim/instrument context.

Produce a proposal when the main output is a recommended PMKNB action, record, or next research task.

Produce no_publish when the package could misidentify the instrument, misstate rules, collapse non-equivalent markets, or present action guidance.

When working inside a PMKNB runner mode, respect runner-owned writes: inner agents should produce the requested output or apply batch for the runner rather than directly committing KNB changes.

PMKNB Runner Notes

When this skill is used from a PMKNB workflow runner:

  • Read world context first, then market state. Market analysis compares world belief against exact instrument rules and current market state.
  • Validate exact venue identity before any forecast: platform, event slug or ticker, market slug or id, condition id, token id when available, exact selection label, outcome, and rules URL or resolution text.
  • Verify the child market creation/listing timestamp and the timestamp of every alleged qualifying occurrence. Pre-creation occurrences are excluded unless the rules explicitly say otherwise.
  • If no exact instrument is available, emit a proposal or no_publish; do not forecast.
  • If the trusted market provider is unavailable, emit a blocked result with reason market_provider_unavailable; do not fabricate price, liquidity, settlement status, or market freshness.
  • Store only selected market snapshots needed to support a forecast, report, or audit trail. Do not retain dense ticks, full orderbooks, or continuous price history.
  • Forecasts are immutable thesis revisions. If writing PMKNB records, write a new forecast and link it with supersedes rather than mutating the old one.
  • Forecast fields should cover thesis type, side, as-of time, current price, fair value or target price, horizon, max entry, exit plan, invalidation, confidence, basis record ids, and basis source ids.
  • If a readable synthesis is needed, use a report or brief product. Do not introduce memo as a product term.
  • Runner-owned writes go to one apply batch for the runner to commit. Do not call knb apply or knb add from inside the analysis.
  • Every meaningful runner-backed package should include one run trace with provider status, instrument id, forecast id if written, snapshot ids, proposal ids, and final status.

Minimum Output Shape

schema: pm_market_analysis.v1
product: forecast | report | brief | proposal | no_publish
pm_event:
  title:
  venue:
  event_identity_status:
instruments:
  - venue:
    identity:
    outcome:
evidence:
  sources: []
  claims: []
  market_snapshot: {}
analysis:
  market_forces: {}
  resolution_oracle_risk: {}
  liquidity_microstructure: {}
  platform_counterparty_risk: {}
conclusion:
  probability_world:
  probability_resolution:
  confidence:
  actionability:
  product_rationale:
validation:
  gates_passed: []
  gates_failed: []
pmknb_write_matrix: []
run_trace: {}

Hard Prohibitions

Do not produce:

  • Buy/sell/hold guidance.
  • Orders, executions, or trading instructions.
  • Dense orderbook history.
  • Private solvency speculation.
  • Title-only instrument matching.
  • Cross-venue equivalence claims without verified identity and rules.
  • Forecasts for unverified instruments.

When uncertain, preserve PMKNB integrity over completeness: mark uncertainty, fail the relevant gate, and choose report, proposal, or no_publish instead of forcing a forecast.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.