agentsclimarketplace

Tofable

Skill treylom/tofu-harness/skills/tofable

Use at the start of any non-trivial task (build / fix / analyze / deliver) to run it under the tofable working discipline — honest decomposition, prompt refinement via /prompt when installed (basic-prompting fallback when not), evidence-first execution, and honest completion. Also use when asked to "apply tofable", "work like fable", or "tofable로 진행".From its SKILL.md

Install
npx -y skills add treylom/tofu-harness --skill tofable

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.6 KB, as published. Nobody here has run it

tofable — working discipline (skill form)

Run the task through 5 checkpoints. Do not skip one because it feels obvious; as you pass each, state it in one line.

0. Decompose honestly

  • Restate the goal in ≤2 sentences: what the requester holds when this is done.
  • Split into verifiable units — each unit names the evidence that will prove it (a command, a diff, a file, a number).
  • Mark what is explicitly NOT in scope (1 line).

1. Refine the prompt (integration point)

  • If a prompt-refinement skill is installed (e.g. /prompt from the prompt-engineering-skills plugin — check your available skills), invoke it on the task statement before executing (/prompt --batch <task>) and adopt the refined structure.
  • If it is NOT installed, fall back to basic prompting: rewrite the task as role + context + constraints + expected output format (4 lines suffice). Never start executing from a bare one-line goal.

2. Execute with evidence

  • After every change, run the verification that unit named (test / build / probe) — before moving on, not batched at the end.
  • Delegating? The first message carries the complete HOW (tools, paths, parallelism, forbidden routes). A delegate's "done" is a claim, not evidence — re-verify after it returns.
  • The same command failed twice? Stop. Diagnose before any retry.

3. Report blockers plainly

  • Blocked = say so immediately: what is blocked, and whose decision unblocks it. Not at the end, not narrated around.
  • Never invent busywork while blocked on someone else's decision.

4. Close honestly

  • Before saying "done": re-run the top-level verification and state its grade — deterministic check / scenario replay / independent evaluator / human review.
  • Any claim with a number ("83 files", "byte-identical") needs a mechanical check in the log (wc / grep -c / diff / shasum).
  • Remaining gaps are listed as gaps; they are never absorbed into "done".

Notes

  • This skill is the proactive twin of the tofable gates (hooks/ in the tofable repo): the gates bounce you when evidence is missing at stop time; this skill makes that evidence exist before the gate ever fires.
  • Gates docs & kill switch: hooks/README.md (FABLE_GATE_OFF=1).

Keep looking

Skills are one crate of 609,856. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.