agentleFS
akshay-diwadkarakshay-diwadkar/first-four-hours/skills/sensemaking/reason/SKILL.md

reason

Skill1 starsChanged 4 days agoMITView on GitHub

Consult at every judgment checkpoint: before committing to an approach, after a failure or direction change, and before presenting a result. Also unblocks decisions, challenges consequential claims, or asks irreducible user-owned questions. Skip only mechanical work with nothing left to judge.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    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:

    decide: no defensible next move exists because goal, evidence, conflict, cause, options, assumptions, or failure remain unresolved: decide …

    line 22
  2. Distinctive:

    No result is approval or proof of correctness.

    line 26

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Reason

Temporary, stateless, and artifact-free: Reason never mutates any state, gives no mutation or domain authority to itself, and outputs only the consultation result, with no score, confidence, or vote. The owning task keeps artifacts, acceptance, completion, and implementation decisions.

Checkpoints

Applicability follows open judgment, uncertainty, and consequence, never task size or file count. Work that chooses an approach, weighs evidence or tradeoffs, or claims an outcome no mechanical check settles must consult Reason before committing to an approach, after a failure or direction change, and before presenting the result. A failed tool, probe, identity check, or seal the work relies on is a failure. Considering consultation, or affirming untested, is not consultation; confidence or one plausible move is no exemption, because owners cannot see their own blind spots. Work with nothing left to judge, such as a rote edit, routine technical repair, or applying a decision-complete plan, skips checkpoints; a plan contradiction returns to the planning owner.

Each consultation is one bounded pass scaled to consequence: the current goal, proposal, or result; its authority and evidence; and the action it could change. On a non-trivial decision, state the strongest opposing argument, test blind spots, and verify every load-bearing assumption. Each turn re-reads your context, so issue the pass's independent reads, searches, and checks together, never skipping one to save calls. It ends with a visible one- or two-line result naming the examined claim or decision, the risk tested, and one disposition: material objection, missing evidence, needed user input, or no material concern; never manufacture criticism or persist the line. Reason explains what to reconsider and why; the owner weighs it and fixes supported gaps. No recursion, panels, or automatic call chains. Each consultation reads Results for what is non-trivial, where the line and results go, and what an open disposition requires. Owners report checkpoints per Checkpoint trace.

Branches

Choose the smallest branch owning the concern by judgment; branches may reclassify internally. Read the branch reference on entry; it holds the user-facing result shape.

  • decide: no defensible next move exists because goal, evidence, conflict, cause, options, assumptions, or failure remain unresolved: decide details.
  • challenge: a concrete proposal or result exists: challenge details, which says when to brief a fresh challenger.
  • ask: progress depends on a material user-owned fact, choice, preference, policy, permission, risk tolerance, scope, authority, or operational rule that evidence and bounded exploration cannot derive: ask details.

No result is approval or proof of correctness.

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.