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.
npx skills add TimboGP/timbogp-marketplace --skill run-experimentgit clone --depth 1 https://github.com/TimboGP/timbogp-marketplaceWrote 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.
[](https://agentmods.dev/skills/timbogp/timbogp-marketplace/run-experiment)<a href="https://agentmods.dev/skills/timbogp/timbogp-marketplace/run-experiment"><img src="https://agentmods.dev/badge/skills/timbogp/timbogp-marketplace/run-experiment/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.
<a href="https://agentmods.dev/skills/timbogp/timbogp-marketplace/run-experiment"><img src="https://agentmods.dev/badge/skills/timbogp/timbogp-marketplace/run-experiment.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00150 | $0.00994 |
| Opus 5 | $0.00075 | $0.00497 |
| Sonnet 5 | $0.00030 | $0.00199 |
| Haiku 4.5 | $0.00015 | $0.00099 |
Grade A, and why
run-experiment 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 12d 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.
How it starts
The opening of the file, as written. The whole thing — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run an experiment
A single trip around the Build-Measure-Learn loop. This skill turns a risky assumption into a falsifiable hypothesis, designs the smallest test that produces real learning, and logs the result. You work as the Analyst guide role (../../reference/roles.md). Read references/build-measure-learn.md before designing.
When to use
The user wants to test an assumption (usually the top one from prioritize-risks), needs help making it testable, or wants to record what an experiment taught them. Customer interviews are a kind of experiment but have their own skill (customer-interview); use this skill for landing-page / smoke / concierge / pricing / channel tests and for the general framing and logging of any experiment.
Core stance
- Maximize speed, learning, and focus together. Drop any one and you get a failure mode: no learning → chasing your tail; no speed → run out of resources; no focus → premature optimization.
- One key metric per experiment. Decide the single number you're moving and ignore the rest for this experiment.
- Smallest thing possible to learn. You rarely need code. Mock-ups, a landing page, a concierge ("Wizard of Oz") MVP, a manual back end, a fake-door button — pick the cheapest artifact that yields a real signal.
- Bold before product/market fit. Pre-fit, you learn most when the outcome is ~50/50, so prefer bold experiments (a different UVP, a different segment) over incremental tweaks. Save micro-optimization for after fit.
Procedure
- Load context. Read
.lean/risks.md(what's riskiest),.lean/PROGRESS.md(stage), and the canvas. Confirm which assumption this experiment attacks. - Write a falsifiable hypothesis. Form:
[specific repeatable action] will [expected measurable outcome]— a statement that can be clearly proven wrong. Make the outcome a concrete number and threshold ("at least 20% of landing-page visitors will enter their email"), not "people will like it." - Pick the single key metric and the success/fail threshold up front (pre-commit, so you can't rationalize a fuzzy result into a "yes").
- Design the smallest build. Choose the cheapest artifact that tests it (see the menu in
references/build-measure-learn.md). Define how you'll measure and the minimum sample/time to trust the signal. - Run / record. When results are in, record the measurement, then the learning: did it validate or refute the hypothesis, and what decision follows (proceed / pivot / iterate / kill)? Capture surprising side-learning loosely as insights to fold into the canvas later.
- Write
.lean/experiments/<id>.mdfromassets/experiment-template.md(<id>likeE1,E2). Update.lean/PROGRESS.md(a dated Journal line). If the result retires or moves a risk, update.lean/risks.md. - Close the loop. Name the next experiment the result implies, and route (back to
prioritize-risksif the riskiest assumption changed,lean-canvasto fold in learning,measure-fitif you're accumulating fit signal).
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 12d ago First seen · 42 lines · 150 tokens per session scan A b43e88c99743
run-experiment is a skill published in the GitHub repository TimboGP/timbogp-marketplace (3 stars, last pushed 2mo ago), licensed MIT. It adds 150 tokens to every session and 994 once invoked, about $0.0007 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.
Other skills, from other repositories
recursive-decomposition
Decompose dense codebase-wide, multi-document, PDF, and aggregation work even when the input fits the context window, following Recursive Language Models (Zhang, Kraska, Khattab, 2025). Use when the user asks to analyse all files, a whole repo, all docs, large PDFs, or to aggregate or multi-hop across scattered…
test-driven-development
Strict red-green-refactor TDD workflow for implementing features, fixing bugs, or changing behavior in Rails applications. Enforces the discipline of writing a failing test before any production code. Use whenever you want to implement with TDD — whether a new feature, a bugfix, a refactor, or any behavior change.
git-commit
Turn the working changes into one or more atomic commits with well-written messages. Use whenever the user runs /git-commit or asks to commit their work, wrap up a feature, or "commit what I have.".
swarm
Run a multi-agent audit of a codebase by spawning specialized parallel subagents (security, performance, tests, architecture, dead-code), then synthesize their findings into a single prioritized action plan. Use this whenever the user runs /swarm, asks to "audit the repo," "review this codebase," "find issues across…
full-security-review
Structured security audit covering injection, auth, secrets, input validation, dependencies, cryptography, and AI/LLM risks. Produces severity-graded findings.
log-correlation
Correlates and troubleshoots logs across OS (Linux/macOS), AWS (CloudWatch, CloudTrail, ALB, Lambda), application (JSON, logfmt), and web servers (Nginx, Apache).