AGENT SKILL

continuous-discovery

暂无简介

License 许可
Apache-2.0
最近更新
2026-11-15
Stars
★ 3
Forks
0

03 SKILL.md 正文

Continuous discovery — interrogate the problem before designing the solution

The kit's own problem-space discovery craft: take a felt need → a validated problem, a measurable outcome, and the riskiest assumption tested — before design shapes any solution. This is the front of the loop (Discover), the partner to skills/design/SKILL.md's solution-space. Continuous discovery is Teresa Torres' practice made workable for a human↔AI pair; it is a kit-original (superpowers has no discovery skill — its brainstorming is solution-space, the kit's design).

Not to be confused with the discovery keystone. skills/using-skills/SKILL.md is skill-discovery ("which skill applies?"). This skill is product discovery ("which problem, and is it the right one?"). Different axis entirely.

<!-- The frontmatter and the discipline markers below are conformance-load-bearing: conformance/orchestrator-loop-wired.sh greps this file for kit-distinctive markers (name: continuous-discovery, discovery partner, outcome over output, opportunity solution tree, riskiest assumption, small bet). Edits that drop or rename them can turn the skill-spine lock RED. -->

When to use

At the front of the loop, before design: when the problem is taken for granted, when a slice's value rests on an assumption no one has tested, or when "what we'll build" is clear but "what outcome it changes, and for whom" is not. Every non-trivial slice deserves a discovery pass — a small one for a well-understood problem, a deliberate one when the problem is fuzzy or the bet is large.

The discovery partner — not the decider

The discovery partner is the kit's framing of the human↔AI relationship here, and it is the load-bearing constraint of this whole skill: the human is the PO. The agent does not decide what to build or declare a problem worth solving. The agent makes continuous discovery accessible and honest: it structures the space, asks the sharp question, names the assumption everyone is gliding past, and designs the cheap test — then hands the judgment back to the human. A partner that quietly starts deciding has stopped being a partner; that is the failure mode this skill exists to prevent. There is no Product seat for the same reason — the human owns the decision; the Orchestrator merely wears the Product hat to invoke this craft.

The flow (the proven spine)

  1. Anchor on the outcome, not the feature — outcome over output. Before anything, ask what customer or business outcome this work would change, and how we'd measure it. Frame the work by the outcome it moves, never by the feature shipped. A slice with no outcome hypothesis is output theatre — surface that out loud. (This is the Definition-of-Ready "success metric / hypothesis" item, made a craft.)
  2. Map the opportunity solution tree. Build the tree with the human: the outcome at the root → the opportunities beneath it (the problems, needs, and desires that, if addressed, move the outcome) → candidate solution ideas → the experiments that would test them. The point is to make the chosen opportunity explicit and compared, not the first idea that arrived. Breadth before depth.
  3. Surface and rank the riskiest assumption. Every chosen opportunity and solution bets on assumptions — desirability (do they want it?), viability (does it work for the business?), feasibility (can we build it?), usability (can they use it?). Name them, then rank the riskiest assumption: the one that, if false, wastes the most work. That is what gets tested first — not the easiest thing to test.
  4. Design the small bet. Design the cheapest experiment that would prove or kill the riskiest assumption before a slice is built — a small bet, not a big-bang study. Continuous, not one-shot: small, frequent tests on a cadence beat a single upfront discovery phase. Define in advance what result would change the decision; a test whose outcome changes nothing is theatre.
  5. Hand off to design. Discovery's terminal state is a validated problem + a measurable outcome + the riskiest assumption tested — handed to skills/design/SKILL.md to shape the solution. This skill never starts a solution; designing is design's job, building is the Engineer's.

What this craft points at (it does not duplicate the infra)

  • Capture the validated problem + outcome in templates/FEATURE-REQUEST-TEMPLATE.md (incl. its success-metric / hypothesis fields).
  • The Definition-of-Ready "success metric / hypothesis" item is the outcome bar this craft feeds; conformance/discovery-complete.sh is the gate that checks the artifact exists.
  • Prove any claim made during discovery with skills/verification/SKILL.md (evidence before assertions — a test result is a claim until you've seen it).
  • Hand the validated problem to skills/design/SKILL.md; the loop continues plan → build → review from there.

Honest ceiling

Discovery quality is un-gateable — there is no CI check that an opportunity tree is good or an assumption well-chosen, just as there is none for design quality. The structural proof is only that this craft is provided, kit-distinctive, and wired to the conductor; the honest enforcement of outcome rigor is the Definition-of-Ready success-metric item and discovery-complete.sh, which this craft feeds. Name that ceiling; never let a green check imply the discovery was actually good.

Rationalizations to refuse

Rationalization Why it fails
"The problem is obvious, just build it." Obvious problems hide the riskiest assumption. Name it and design a small bet before building — that is the cheap insurance.
"We agreed on the feature, discovery is done." A feature is an output. Until you can state the outcome over output and how you'd measure it, discovery is not done.
"I'll run discovery and tell the human what to build." You are the discovery partner, not the decider. The human is the PO; structure and surface, then hand back the judgment.
"Let's just pick the first solution." An opportunity solution tree exists so the chosen opportunity is compared, not defaulted. Breadth before depth.
"We'll validate it after we ship." Continuous discovery tests the riskiest assumption with a small bet before the build — post-hoc validation is the waste it prevents.

Terminal state

A validated problem, a measurable outcome (outcome over output), an explicit opportunity solution tree, and the riskiest assumption tested with a small bet — captured in the feature request and handed to skills/design/SKILL.md. The human made every call; the discovery partner kept it honest. This skill never starts implementation.

04 同仓库更多 Skills

SeaBrad72
SeaBrad72/sparkwright

inception-from-code

Use when adopting the kit into an EXISTING repository (brownfield), to draft the Inception artifacts from the code instead of from blank templates. The agent reads the repository and writes five drafts, only under docs/inception-draft/, each tagged by how each fact is known; a person reviews them and moves what is right into place. Agents propose, humans ratify.

★ 3
SeaBrad72
SeaBrad72/sparkwright

worktrees

Use BEFORE fanning out parallel Engineers or starting isolated feature work — the kit's own isolation skill (replaces, does not depend on, superpowers using-git-worktrees). The isolation craft, wired to the Orchestrator seat — detect-existing-first, native tools first, the kit's disjoint-set parallel-safety rule, and conflict-safe integration.

★ 3
SeaBrad72
SeaBrad72/sparkwright

verification

暂无简介

★ 3
SeaBrad72
SeaBrad72/sparkwright

using-skills

Use at the START of any task, before any response or action — the kit's own discovery keystone (replaces, does not depend on, superpowers using-superpowers). How the kit's skills are found and invoked (invoke by reading), the discipline of checking for a relevant skill before acting, instruction priority, and the index of the kit's own spine skills.

★ 3
SeaBrad72
SeaBrad72/sparkwright

tdd

Use when implementing any feature, bug fix, refactor, or behaviour change — BEFORE writing production code. The kit's own test-driven-development skill (replaces, does not depend on, superpowers test-driven-development). Red-green-refactor framed as the kit's non-vacuity law, plus coverage, testing-pyramid, and AI-eval disciplines.

★ 3
SeaBrad72
SeaBrad72/sparkwright

review

暂无简介

★ 3