agentleFS
iangogenticiangogentic/venture-lab/.claude/skills/contradiction-analysis/SKILL.md

contradiction-analysis

Skill0 starsChanged 4 months agoMITView on GitHub

Search for evidence against each opportunity.

4 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

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    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:

    Every stage before this one built the case *for*.

    line 37
  2. Distinctive:

    No verdict: Do not say whether the opportunity should proceed.

    line 73
  3. Distinctive:

    No weighing: Do not net counter-evidence against the case for, do not say which finding matters most, do not write "on balance".

    line 74
  4. Distinctive:

    If the search turns up nothing, return nothing — and record what you searched for.

    line 87

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Contradiction analysis

Go hunting for evidence against each opportunity below. Return what you find. Do not weigh it, and do not conclude anything.

The question

${question}

The opportunities

${opportunities}

The market analyses

${market}

The competition analyses

${competition}

Why this stage exists

Every stage before this one built the case for. Left alone, a pipeline of summarisers converges on a coherent story, and coherence is not truth. Your job is adversarial: read all of this as someone who wants the opportunity to be wrong, and go looking for what would show it.

One analysis per opportunity, linked by its real id. Argue against that specific opportunity — not against the category, not against the market in general.

What to search for

  • Failed startups — someone tried this, or close enough to this, and did not make it. What did they build, for whom, and what is known about why it ended?
  • Negative reviews — users of the existing answers saying they do not work, or that they work well enough that nothing better is needed. Both are counter-evidence.
  • Abandonment — people who adopted something for this problem and stopped. Churn, deprecated internal tools, a workflow that reverted to the spreadsheet. Abandonment is stronger counter-evidence than absence, because someone already validated the idea and it still did not hold.
  • Incumbent strength — an existing player well placed to absorb this the moment it matters: distribution, the data, the contract, the relationship, the roadmap adjacency.
  • Market risk — structural conditions working against it: budgets that do not exist, procurement that outlasts a startup, regulation, a platform dependency that could close, a buyer with no incentive to act, a segment too small to reach affordably.

Mine the supplied market and competition analyses as well as the opportunities. The strongest counter-evidence is often already sitting in them unflagged: an adequate substitute, a switching cost nobody can pay, a buyer who is not the sufferer, a size estimate that could not be established. Lift those out and state them as what they are.

Grade severity by what it would do to the decision, not by how uncomfortable it feels: minor where it colours the picture, material where the decision should be framed or conditioned differently, blocking where a decision cannot safely be taken until it is settled. Use blocking sparingly and mean it — it stops real work.

Evidence only

  • No verdict. Do not say whether the opportunity should proceed.
  • No weighing. Do not net counter-evidence against the case for, do not say which finding matters most, do not write "on balance".
  • No conclusion, no recommendation, no mitigation. "This could be addressed by…" is the decision stage's work, and offering it here lets the search stop early — as soon as the story resolves, you stop looking, which is exactly the failure this stage prevents.

Each finding is one observation, stated as an observation, attributed where it can be attributed. Where it came from a supplied artifact, cite that artifact by its real id. Where you cannot attribute it, say so rather than dressing it up; an unattributable objection is still a lead, but it must not look like a fact.

Finding nothing is a real result

If the search turns up nothing, return nothing — and record what you searched for. That list is what makes an empty result interpretable instead of indistinguishable from not having looked. Record it either way, whatever you found.

Do not manufacture objections. Padding this stage is more damaging than padding any other: a decision stage that learns to discount your findings stops reading them, and then the one real objection arrives with the noise.

Failure modes

  • Restating an upstream caveat as though it were a discovery.
  • A named failed company you are not certain existed, or a cause of death you inferred and reported as known.
  • Generic risk boilerplate ("the market is competitive", "execution is hard") that would apply to any opportunity.
  • Grading everything material to look balanced, or upgrading to blocking for effect.
  • Softening a genuinely blocking finding into a hedge.
  • Any sentence that adds up the evidence, recommends, or resolves.

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.