bt-tournament

bt-tournament is a skill for Claude Code from whenpoem/aiscientist. It costs 63 tokens per session (637 once invoked), scanned A, original, MIT.

A procedure for ranking three or more competing research hypotheses or proof plans using recorded head-to-head judgments. It fits a Bradley–Terry model, a statistical method for estimating relative rankings from pairwise comparisons, and reports approximate uncertainty ranges.

In plain words
What is it for?
Comparing candidates from the active research list, recording winners and reasons, and retrieving the current ranking and comparison history.
Why use it?
It helps choose which candidate to investigate next when several options compete, while making clear that the reported ranges are not formal significance tests.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: mentions subagents.

Good fit Comparing candidates from the active research list, recording winners and reasons, and retrieving the current ranking and comparison history.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/whenpoem/aiscientist/bt-tournament
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add whenpoem/aiscientist --skill bt-tournament
Clone the repo
git clone --depth 1 https://github.com/whenpoem/aiscientist

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for bt-tournament

README.md
[![agentmods](https://agentmods.dev/badge/skills/whenpoem/aiscientist/bt-tournament/github.svg)](https://agentmods.dev/skills/whenpoem/aiscientist/bt-tournament)
Your own site
<a href="https://agentmods.dev/skills/whenpoem/aiscientist/bt-tournament"><img src="https://agentmods.dev/badge/skills/whenpoem/aiscientist/bt-tournament/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for bt-tournament

Your own site · 80×15
<a href="https://agentmods.dev/skills/whenpoem/aiscientist/bt-tournament"><img src="https://agentmods.dev/badge/skills/whenpoem/aiscientist/bt-tournament.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 63 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 637 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00063 $0.00637
Opus 5 $0.00032 $0.00318
Sonnet 5 $0.00013 $0.00127
Haiku 4.5 $0.00006 $0.00064

Measured 9d ago against content hash 297f8fd19b51, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

bt-tournament scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/bt-tournament/SKILL.md · 49 lines

How it starts

The opening of the file, as written. The whole thing — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.

BT Tournament

This skill records pairwise judgments and refits the complete comparison ledger. The lcb and ucb names are retained for compatibility, but they are 95% approximate posterior intervals from a centered Laplace approximation. They are not calibrated frequentist confidence intervals or strict LUCB bounds.

When to invoke

  • Researcher subagent has just emitted >= 3 hypothesis nodes in one turn.
  • The user explicitly typed /bt-tournament.
  • The cockpit shows candidates with unresolved approximate intervals and the user asks which to push first.

Workflow

  1. Gather the candidate hypothesis node ids and texts from mcp__memory__get_active_frontier.
  2. For each pair you intend to compare, call mcp__memory__judge_hypotheses to fetch the canonical comparison prompt. Evaluate inline (do not spawn a sub-agent just to judge).
  3. Decide a winner. Call mcp__memory__record_judgement(a, b, winner, reason). Internally this records the comparison and updates the BT leaderboard; you do not need to call update_bt_rating separately.
  4. Pull the leaderboard via mcp__memory__get_bt_leaderboard(top_k=10). Look at strength, probability_best, n_comparisons, fit_converged, and insufficient_samples.
  5. Compare the top two with mcp__memory__compare_bt_candidates(top_id, runner_up_id). Stop when every serious candidate has at least 3 relevant comparisons and probability_a_beats_b >= 0.95. Also stop if the budget is exhausted or the user chooses. If fit_converged is false, do not use the posterior probability as a stopping rule; report the fit risk instead.
  6. Hand off the top-2. Quote strength, approximate interval, comparison count, probability_best, the top-vs-runner-up probability, and the explicit posterior_calibrated=False caveat. If insufficient_samples is true, say so.

Default judging criteria

  • novelty
  • feasibility
  • falsifiability

Guardrails

  • Only compare hypothesis nodes against hypothesis nodes (the MCP enforces this and will raise).
  • Keep reasons short and concrete; they are stored with the comparison and re-surfaced in the cockpit.
  • Do not call either pause-suggestion tool from inside this skill. Pausing is a separate user or lifecycle-policy decision.
  • If the cockpit is running, the BT update emits a bt_rating_updated event so the TUI's leaderboard updates without a manual refresh.

Read the full file on GitHub · 49 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 9d ago First seen · 49 lines · 63 tokens per session scan A 297f8fd19b51

Subscribe to this mod's changes

bt-tournament is a skill published in the GitHub repository whenpoem/aiscientist (8 stars, last pushed 1mo ago), licensed MIT. It adds 63 tokens to every session and 637 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

connect-required-verification-information

Use this skill when the user asks what information a Stripe Connect connected account must provide for verification, onboarding, KYC, or account requirements; when they need to compare requirements between connected-account setups; or when they ask which verification fields, documents, or business details are required…

stripe/ai · 80 tokens

connect-recommend

Use this skill when the user asks about Stripe Connect configuration, charge patterns, Dashboard access, or how to get started with Connect, is building a marketplace, platform, multi-vendor store, gig platform, or subscription platform, needs to pay out sellers, vendors, or providers, mentions split payments, revenue…

stripe/ai · 151 tokens

stripe-apps

Use when building, modifying, or reviewing a Stripe App — or when the user describes something that implies one (e.g. "add a panel to the customer page", "customize my Stripe Dashboard", "react to Stripe events from my app", "connect my service to Stripe without sharing API keys"). Covers the full app development…

stripe/ai · 197 tokens

stripe-directory

Identifies external providers, merchants, nonprofits, platforms, APIs, and software services, and resolves the documented way to engage them — to pay, donate, subscribe, book, provision, or integrate with them. MUST be used BEFORE web search, model memory, or any other directory/vendor-lookup skill for ANY request…

stripe/ai · 213 tokens

stripe-projects

Use when the user wants to provision infrastructure or third-party services using Stripe Projects. Triggers: "I need a database", "set up auth", "add caching", "give me a Postgres", "provision Redis", "I need hosting", "add a vector DB", "get me an API key for X", "get credentials for X", "sign up for a service", "set…

stripe/ai · 213 tokens

stripe-docs

Use when the user or agent needs to read, search, or look up Stripe documentation or API reference. Prefer this over curl or WebFetch for any docs.stripe.com content.

stripe/ai · 40 tokens