agentleFS
q2522879285-sourceq2522879285-source/red-monkey-reasoning-kit/skills/bidirectional-steelman/SKILL.md

bidirectional-steelman

Skill2 starsChanged 2 months agoMITView on GitHub

Apply a bidirectional steelman pass to complex, contestable judgments, decisions, strategies, architecture choices, and recommendations. Use when the user requests a two-sided challenge or when credible alternatives combine with stakes, uncertainty, competing objectives, or conclusion-flipping variables. Skip simple factual lookup, translation, formatting, fixed-scope execution, and deterministic tasks that are directly verifiable.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 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.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run the full pass when any condition holds:

    line 12
  2. Distinctive:

    If one unavailable user fact can realistically flip the result and cannot be inferred or verified, ask exactly one decisive question.

    line 26
  3. Very specific:

    Use $socratic-problem-framing first only when the decision itself is not yet clear.

    line 30
  4. Very specific:

    Use $fact-claim-audit when the two sides depend on disputed or current facts.

    line 31
  5. Distinctive:

    Do not create fake balance: unequal evidence may justify an unequal conclusion.

    line 33
  6. Distinctive:

    Keep the pass backstage by default.

    line 37

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Bidirectional Steelman

Apply this as a reasoning layer around the active domain skill. The domain skill still owns facts, calculations, implementation, and professional standards; this skill owns the contest between plausible conclusions.

Route

Run the full pass when any condition holds:

  1. The user explicitly asks for steelmanning, counterarguments, a two-sided debate, or a challenge to their current view.
  2. A judgment, decision, strategy, architecture choice, or recommendation has at least two signals: credible alternatives; meaningful cost, risk, lock-in, or opportunity cost; material uncertainty or hidden assumptions; multiple stakeholders, objectives, or time horizons; or a variable that could reverse the recommendation.
  3. A high-impact recommendation is contestable and has a credible alternative, even when the user presents only one side.

Complexity is about decision structure, not request length. Skip facts, direct transformations, fixed instructions, and deterministic implementation or bug fixes settled by inspection and tests.

Required Pass

  1. Restate the desired outcome, disputed claim or choice, constraints, observed facts, assumptions, and inferred preferences.
  2. Build the strongest evidence-based case for the leading direction, including causal logic, success conditions, and avoided costs.
  3. Build the strongest credible case against it or for the best alternative using the same evidence standard. Prefer a real conditional or third option over a false binary.
  4. Name the crux: facts, causality, values, risk tolerance, timing, reversibility, or resources. Identify the smallest variables that could change the conclusion and state evidence asymmetry plainly.
  5. If one unavailable user fact can realistically flip the result and cannot be inferred or verified, ask exactly one decisive question. Otherwise give the current judgment, why it wins, confidence when useful, and the next action.

Package Handoffs

  • Use $socratic-problem-framing first only when the decision itself is not yet clear.
  • Use $fact-claim-audit when the two sides depend on disputed or current facts.
  • A plan produced by $first-principles-rebuild, $cross-domain-mechanism-transfer, or $expert-perspective-panel may receive this pass only when a credible alternative can materially change the recommendation.
  • Do not create fake balance: unequal evidence may justify an unequal conclusion.

Delivery

Keep the pass backstage by default. Deliver the conclusion, decisive crux, strongest material caveat, and next action. Show the full two-sided structure only when the user asks for it or the trade-off itself is the deliverable.

Completion Gate

The real problem and constraints are fair; both the leading case and strongest credible opposition were considered; the crux and flip variables are explicit; uncertainty is visible; and the response ends with a usable judgment and action or one decisive question.

Source Status

This skill predates the eight-card pack and is included as its contest layer. Its workflow is a reasoning protocol, not proof that two-sided analysis always improves a decision.

<!-- integration-watermark: 红猴子整合 -->

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.