Skip to content
Agentmart
Submit a tool

design

When intent or consequential choices are unclear — resolve goals, boundaries, contracts, and an approved direction before building.

Skill by bryan824

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/bryan824-design

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

When intent or consequential choices are unclear — resolve goals, boundaries, contracts, and an approved direction before building.

Use cases: design, productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/design

What's inside

Skill name
design
Description
When intent or consequential choices are unclear — resolve goals, boundaries, contracts, and an approved direction before building.
License
MIT
Instructions
46 lines in SKILL.md · view on GitHub

Files (2)

  • SKILL.md
  • DECISIONS.md · references/

SKILL.md (first lines)

# Design

Find the real need before committing to a solution. A clear bounded request can
already supply approved intent; do not turn it into another interview.

- Read what the repository can answer. Separate discoverable facts, user-owned
  choices, and knowledge held by an absent stakeholder. Research and delegated
  evidence inform decisions; they do not authorize them.
- Ask about consequential uncertainty: outcome, audience, non-goals, constraints,
  data and failure behavior, acceptance, and hard-to-reverse choices. Use concrete
  scenarios to expose ambiguous terms or contradictions; check claimed behavior
  against the code before treating it as a fact. Challenge a fashionable solution
  when a simpler one solves the actual problem.
- Bundle dependency-ready questions into a short round. Include a grounded
  recommendation, worded so a plain yes accepts it, and meaningful alternatives;
  ask separately when one answer would reframe the rest. Do not demand arbitrary
  option counts or explore every hypothetical branch. Block only work that depends on an unresolved answer.
- If someone else holds the answer, establish the recipient and needed outcome,
  then prepare a focused questionnaire. Each question should stand alone, lead
  with what matters most, and welcome partial answers or uncertainty. Do not
  invent the absent person's preference.
- State the proposed outcome, exclusions, contracts, test seams (what each
  catches and misses), assumptions and consequential trade-offs clearly enough
  to approve. Approval means the user
  agreed to this direction in context; no magic word is needed, and silence is
  not approval. A material expansion needs fresh approval. If no decision-maker
  is available, report the blocked choice rather than guessing through it.
- Prefer small verifiable, reversible outcomes. Probe load-bearing assumptions
  before designing around them; name what would invalidate the direction.
  Compare architecture or build an authorized prototype when evidence would
  settle the choice more cheaply than discussion.

For dependent choices spanning sessions, use [Decision frontier](references/DECISIONS.md)
inside the task's existing working notes (one ignored `context/<effort>.md`),
never the handoff file. When continuity needs it and writing is authorized,
capture settled constraints, explicit no-s, ordering/numeric requirements and their
reasons as they land, not only in a final summary. Put agreed terms beside the
contract they clarify; a glossary is not a spec. No per-question record, glossary,
or planning artifact is required merely to finish an interview.

Deliver: approved intent and boundaries, acceptance/contract decisions, unresolved
prerequisites and their owners. Proceed within clear implementation authority;
use a plan when risk, dependencies or handoff need one. `plan` and `implement`
defer to these approval rules.

From bryan824/kirin-pi, MIT.

More from bryan824/kirin-pi

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 13, 2026 (16 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from bryan824

Cite this page

design skill. Agent Mart. https://agentmart.dev/skills/bryan824-design (last updated 2026-10-24)