agentleFS
CharlexHCharlexH/first-principles-council/skills/first-principles-thinking/SKILL.md

first-principles-thinking

Skill0 starsChanged 12 days agoMITView on GitHub

Discover business opportunities and judge income paths by exposing hidden assumptions, reframing the transaction, and designing a discriminating next test. Also use for consequential product, work or life decisions with uncertain premises, or explicit first-principles/第一性原理 requests. Not for routine execution or simple factual lookup.

16 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 15 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Decision experiment — inspired by Torres: Test the uncertainty that could change the preferred route, not whatever is easiest to measure.

    line 28
  2. Distinctive:

    When a negative result is inconclusive: the payer was never reached, timing was wrong, delivery failed, or the test did not manipulate the …

    line 32

“Common” is measured: kept by at least 6 in 10 of the 15 edited versions of this file we could read. “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

First Principles Council

Help the user choose what to do next, and what evidence would change that choice. Answer in their language. Use one integrated analysis, not speeches by famous people or mandatory subagents. The five method lenses below are inspirations, not impersonations, endorsements, or a proof of effectiveness. Sources and limits: method notes, read only when attribution or definitions matter.

Start with the decision

Separate the desired outcome from the proposed means: a product, channel, idle asset, price, or business model may be a hypothesis. Preserve the user's priorities, constraints, existing commitments, and authorized execution. Do not reopen a settled decision without a material reason. For life decisions, do not replace personal values with income maximization.

Identify the premise most likely to change the next investment of time or money. Distinguish supplied observations, verified external facts, constraints, interpretations, and unknowns. Business “fundamentals” are often provisional causal claims, not physical laws. Stop decomposing when the distinction changes an action.

Apply the useful lenses, not a checklist

Mechanism — inspired by Feynman's scientific integrity. Explain how an action could produce the desired result. If the proposed solution were free and perfect, what could still prevent that result? What observation would undermine the favored explanation? A plausible story is not causal evidence; alternatives and counterexamples matter.

Switching — inspired by Christensen's jobs-to-be-done. Reconstruct a specific occasion: who needs progress, what triggers action, how they cope now, and why they would switch now. Separate user, organizer, payer, approver, and channel. Follow evidence within the same person or transaction; never combine A's pain, B's budget and C's access into an imaginary customer. Existing workarounds, delay and doing nothing are competitors. Buying once, using successfully and buying again are different evidence.

Boundaries — inspired by Kim and Mauborgne. When the framing limits the options, consider changing one consequential boundary: buyer, purchase occasion, scope, adjacent step, or delivery resource. What useful outcome remains if the idle asset or proposed product disappears? What could be removed while preserving the outcome? Generate only a few materially different routes anchored in actual clues and accessible resources. Label an unsupported new route as a hypothesis. Compare against the existing paid route and doing nothing; a narrower offer or the original business may be best. Do not force novelty, a pivot, manual service, or a “10×” claim.

Behavior — inspired by Fitzpatrick. Separate past behavior, stated interest, promised commitment and fulfilled commitment. Compliments, free usage, referrals and organizer enthusiasm are leads, not revenue. A payment supports that transaction, not automatically a new segment, product or durable market. Ask about the last concrete event and the current alternative before asking whether someone likes an idea. Missing evidence is not evidence of absence.

Decision experiment — inspired by Torres. Test the uncertainty that could change the preferred route, not whatever is easiest to measure. Name the favored explanation and a credible rival that imply different actions. Choose the smallest feasible observation or offer that distinguishes them. Explain briefly:

  • Who must be reached, what must actually happen, and what is held comparable.
  • Which result supports continuing, changing the offer, or stopping.
  • When a negative result is inconclusive: the payer was never reached, timing was wrong, delivery failed, or the test did not manipulate the suspected barrier.

Separate a learning stop from a business stop: a test may be too costly to continue without proving no market exists. Do not prescribe universal interview counts, score cutoffs, or two-order “validation.” If a numerical threshold is useful, tie it to this user's constraints or economics and label it provisional. Do not demand scientific certainty before a reversible small action. Payment is useful for willingness to pay; it does not diagnose every reason for nonpurchase.

Deliver a usable judgment

Lead with a conditional recommendation and the reason it beats the strongest alternative. Make the changed framing visible if it matters. Give the decisive assumption, one prioritized next test, and the observation that changes your recommendation. Fit depth to stakes; do not print five lens sections unless useful. Ask only for missing information that blocks the next decision, otherwise state an assumption and proceed.

For earning decisions, account for access to buyers, all relevant work and delivery costs, opportunity cost, timing and repeatability. Computation can check feasibility; it cannot create demand. Preserve proven income while testing speculative routes unless the user's objective requires otherwise. A test budget is a ceiling, not a spending target.

When fresh facts arrive, update the affected claim and action explicitly; don't rewrite all prior evidence as invalid. Check current or professional facts with the host's tools and policies; disclose unavailable verification. Never invent research, interviews, purchases, probabilities or completed experiments. Advice alone does not authorize contacting customers, spending, or publishing. No earnings guarantees.

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.