agentleFS
KaidevLabsKaidevLabs/Skills/skills/eureka/SKILL.md

eureka

Skill1 starsChanged 34 days agoMITView on GitHub

Discovery protocol for analyzing a problem with the owner before planning or code — use when starting a new feature, bug, or problem that needs decisions settled, or before an Epic is planned.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. ArangoGutierrez/claude-toolkit★ 4same name
    2. PythonShe/Repertoire★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository7 other files it gives its agents.

The file

Files
SKILL.md

Eureka

The discovery phase — analyzing a problem with the owner before any plan or work. Use whenever a problem is big or ambiguous enough that deciding wrong costs a session, and always before an Epic is planned.

Protocol

  • Evidence before questions: verify the current state with evidence first — grep, DB, tests, logs. Only ask what the evidence cannot answer. Every current-state claim in the resulting Epic carries its evidence.
  • One question at a time. The owner decides; the agent never stands in for the owner's side of the exchange — an agent that answers its own questions has broken the protocol.
  • Grill before settling: challenge every decision before it lands in the do-not-revisit table — what does it complicate, what does it rule out, what would prove it wrong. Settled is do-not-revisit; pre-settle is the last cheap place to be wrong out loud.
  • Record settled decisions immediately: each one lands in the Epic's Settled decisions table (A# | decision | rationale | date), marked "do not revisit" — recorded at once in the draft while discovery runs, via issue-body edits (API writes) once the Epic is published. A decision without its rationale is an invitation to relitigate.
  • Track what is not yet specifiable: questions you can tell are coming but cannot yet state sharply go to Not yet specified. The test is whether the question is sharp, not whether it is answerable — sharp questions become steps, the rest waits.
  • Close with Out of scope (work ruled beyond the goal — scope, not sharpness, lands it there) and a Definition of done checklist whose items are verifiable.
  • Close the branches before stopping: for every settled decision, one level out — what hangs on it. Each branch must end recorded: settled (the Epic's Settled decisions table), asked in the session, parked deliberately as a sharp open question (a line in the Epic's Open branches), or fog (Not yet specified). An opened branch may not silently evaporate — "the way is clear" is something the owner can verify at Epic review, not an agent assertion.

Close

Stop when the way is clear — the discovery closes by drafting the Epic body (Goal, Settled decisions, Current state, Definition of done, Not yet specified, Out of scope; plus Open branches when sharp questions are deliberately parked). Execution work hands the draft to planning (the plans skill) to add the execution order, the executor rules, and the step Tasks with their wiring, per the tracker doc. Design-only efforts — decisions, no code — publish as a single Epic with no step Tasks; the owner's ratification closes it SETTLED. The pull to start doing the work is the signal the discovery is done.

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.