agentleFS
iangogenticiangogentic/venture-lab/.claude/skills/decision/SKILL.md

decision

Skill0 starsChanged 4 months agoMITView on GitHub

Choose build, reject or wait for each opportunity.

19 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use explore, brief and 5 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • genomewalkerSkill

    explore

    RLM-style recursive memory exploration - dynamically navigate the memory graph

    Also in 3 projects that use this

  • AnthropicSkill

    brief

    Generate contextual briefings for legal work — daily summary, topic research, or incident response. Use when starting your day and need a scan of legal-relevant items across email, calendar, and contracts, when researching a specific legal question across internal sources, or when a developing situation (data breach, litigation threat, regulatory inquiry) needs rapid context.

    Also in 3 projects that use this

  • BlockSkill

    plan

    Draft a TDD, PRD, or lightweight plan in the conversation without committing it.

    Also in 4 projects that use this

  • SentrySkill

    spec

    Produce a verified spec — problem, desired outcome, and acceptance criteria — before building. Use when given a GitHub/Linear issue link, when starting from a fuzzy in-conversation idea, or asked to "figure out what to do" / "scope this" / "what should I build" — whenever the *what* isn't yet pinned down. Grills the ambiguities, checks the premise against the code, and produces acceptance criteria you sign off on — then hands to design-first.

    Also in 3 projects that use this

  • OpenAISkill

    audit

    Audit or critique a product flow, journey, workflow, funnel, onboarding path, checkout path, settings path, screen, or multi-step product experience by capturing screenshots first, then reporting UX, design, and accessibility findings inline from that evidence. Use Figma only when the user explicitly asks for a board. Use when the user asks to audit, review, critique, inspect, assess, analyze, evaluate, or give feedback on a product experience.

    Also in 3 projects that use this

  • SentrySkill

    review

    Fixture review skill.

    Also in 4 projects that use this

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every stage before this one was forbidden from netting its findings against anything.

    line 61
  2. Very specific:

    A crowded field is not by itself a reject: Read competitors together with switching_costs and substitutes.

    line 78
  3. Very specific:

    differentiation is a claim, not a finding: Treat an entry there as something still to be validated unless something in the evidence …

    line 85
  4. Very specific:

    Read the basis and decide what the gap actually is:

    line 98
  5. Very specific:

    If a contradiction analysis found nothing, note that you saw it — and note whether its searched_for list was thorough enough for the …

    line 133

“Common” is measured: kept by at least 6 in 10 of the 18 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

Decide

Rule on each opportunity below — build, reject or wait — with reasoning explicit enough that someone can re-judge the call later and see exactly where they disagree with you.

The question

${question}

The opportunities

${opportunities}

The market analyses

${market}

The competition analyses

${competition}

The contradiction analyses

${contradictions}

One verdict per opportunity

Exactly one, linked to the opportunity by its real id. There is no "build with reservations": that is build, plus a next validation step that says what you are watching.

  • build — pursue it now. The case holds and the remaining risks are ones you are knowingly accepting.
  • reject — do not pursue, and do not expect that to change. A judgement about the idea: the pain is not worth solving, or this is not the way to solve it.
  • wait — plausible, but something has to change or be learned first. Name what, and name what would tell you it has happened. A wait with no trigger is avoidance wearing a verdict.

The distinction most often collapsed is reject versus wait. Rejecting because you do not know enough closes a question nobody actually examined; waiting on an idea that is plainly bad burns a research cycle. Be clear which one you mean. And do not use wait as the safe middle: if every opportunity comes back wait, you have decided nothing.

This is the only stage that weighs

Every stage before this one was forbidden from netting its findings against anything. The market analysis made the case for and did not discount it; the competition analysis described the field and did not conclude from it; the contradiction analysis returned findings and no verdict. The trade-off is unresolved on purpose, and it is yours. Say plainly what you traded against what — the weighing has to be visible in your reasons, not just implied by the verdict you landed on.

How to weigh the four readings

The dimensions are not interchangeable. They do not add up to a score, and a strong reading on one does not buy down a weak reading on another.

  • A strong market case does not rescue an opportunity with a blocking contradiction. A large, well-derived market makes a blocked opportunity a bigger miss, not a smaller risk. If a finding is graded blocking, the size of the prize is not the answer to it — either a reason states plainly why that finding does not bind this decision, or the verdict is wait with the finding as the trigger.
  • A crowded field is not by itself a reject. Read competitors together with switching_costs and substitutes. Many incumbents with low switching costs means customers can move — the field is contested but not closed, and the honest verdict is usually wait on evidence that a wedge exists, not reject. High switching costs, or a moats entry that a new entrant has no route around, is what turns a crowded field into a genuine reject. Note also that substitutes are usually the real competition: a spreadsheet that is good enough is a harder opponent than a funded company.
  • differentiation is a claim, not a finding. Treat an entry there as something still to be validated unless something in the evidence actually supports it.

An unsized market is missing information, not a negative signal

A SizeEstimate can legitimately carry no number. Its basis is mandatory either way: with an amount it is the derivation, without one it is why the market could not be sized. The previous stage was built to make that refusal expressible so that nobody would have to invent a figure.

So do not read an absent SAM or SOM as a small market. Treating "we could not size this" as evidence against would punish exactly the honesty the stage was designed for, and it would teach the pipeline that a fabricated number scores better than an admitted gap. Read the basis and decide what the gap actually is:

  • Unsized because the buyer segment is not yet identifiable → that is your biggest unknown and probably a wait with sizing as the next validation step.
  • Unsized because the figure exists but was not to hand → it is a cheap next validation step and should barely move the verdict.
  • Sized with a basis that is a chain of guesses → weaker than it looks. A number with a thin derivation deserves less weight than an honest refusal.

The same reading applies to pricing, to assumptions, and to unknowns — an empty unknowns list is rarely honest, and is itself worth noticing. And check the buyer against the budget_owner: when they differ, someone has to be persuaded who does not feel the pain, and that is a real cost the size figure does not capture.

Reasons must be traceable

Each reason points at something in the artifacts you were given — an opportunity's why now, a size estimate that could not be established, a switching cost, a specific piece of counter-evidence — and cites it by real id where it has one. A reason resting on a fact that appears in none of these artifacts is not a reason; it is a preference.

Give the two or three things that actually drove the call, strongest first, not everything true about the opportunity. Include what argued the other way and why it lost. A list of only supporting points is advocacy, and it gives a later reader nothing to check.

The counter-evidence has to show up

The contradiction analyses were produced deliberately, without weighing, so that the weighing happens here and is visible.

  • A blocking finding pushes hard away from build. Building over one is permitted, but only if a reason states plainly why that finding does not bind this decision. Otherwise the honest verdict is wait.
  • material findings do not block, but should appear somewhere: in a reason, in the next validation step, or in an explicit statement that the call holds either way.
  • If a contradiction analysis found nothing, note that you saw it — and note whether its searched_for list was thorough enough for the emptiness to mean anything. An empty result from a shallow search is not reassurance.

Confidence, unknown, next step

Required for every verdict, including reject, and including a build you are sure of.

  • Confidence grades this judgement, not the idea's attractiveness. Anchor it to how much the artifacts actually support the call. Thin, sparse evidence with a clear direction is a low-confidence build, and saying so is what makes the call reversible. Four readings that agree do not raise confidence if all four rest on the same one piece of evidence — check whether they are independent before you let them accumulate.
  • Biggest unknown is required even when you are confident — especially then. A confident verdict that cannot name what it is still most likely to be wrong about has stopped reasoning. Name the specific thing, not "market dynamics".
  • Next validation step is the single cheapest action that would move the confidence: who to talk to, what to look up, what to try. Days, not quarters. It applies to reject as much as build — what would make you revisit this, and what do you tell whoever asked the question?

Failure modes

  • A verdict that does not follow from its own stated reasons.
  • Netting the four readings into an implied score, or letting market size answer an objection that is not about size.
  • Reading an unsized market, or an honest basis, as evidence against.
  • Rejecting on the mere presence of competitors, without reference to switching costs, substitutes or moats.
  • wait used to avoid deciding, or with no trigger attached.
  • Building on everything, or rejecting everything, which is the same failure twice.
  • Reasoning that appeals to facts none of the supplied artifacts contain.
  • Silence about the strongest argument against the verdict you chose.
  • Confidence set by how good the idea sounds rather than by how much stands behind it.
  • A biggest unknown so vague that no evidence could ever resolve it.

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.