idea-to-build
Start a local Hermes Idea-to-Build session after the human explicitly invokes this skill.
Skill by eiei114
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/eiei114-idea-to-build
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.
About
Reviews
Write a reviewNo 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 and code. Checked Oct 26, 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/idea-to-build
What's inside
- Skill name
idea-to-build- Description
- Start a local Hermes Idea-to-Build session after the human explicitly invokes this skill. Use ONLY when the user runs /skill:idea-to-build (never auto-detect). Keep Multica Project and Spine binding deferred until build_handoff.
- License
- MIT
- Compatibility
- Requires multica CLI (user token), npm run build, and pi-multica-spine scripts on PATH or repo root.
- Instructions
- 121 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Idea to Build Human-initiated entry into the **Hermes Idea-to-Build** workflow on the **sandbox canary** lane. ## When to use - The user explicitly typed **`/skill:idea-to-build`** (or `/skill:idea-to-build <idea>`). - Do **not** load this skill from casual conversation alone. ## Your job 1. Confirm the skill was invoked deliberately. 2. If the rough idea is not already in the message after the command, ask once: **「どんなアイデアを作りたいですか?」** 3. Do not debate the idea at length — capture it verbatim (light formatting OK). 4. Bootstrap the workflow only (see below). Do not advance a campaign during entry. 5. After explicit human approval, advance the local lane one stage at a time until `promotion_ready`. Report the explicit `visualTarget` when supplied. Do not report a Multica parent or Project before promotion preflight succeeds. ## Bootstrap (live) From the **pi-multica-spine package root** (repo clone or installed npm package): ```bash npm run build node scripts/workflow-idea-entry.mjs --rough-idea "<ROUGH_IDEA>" --execute --json ``` For an iOS/game or other visual product, pass the target explicitly. Never infer it from keywords: ```bash node scripts/workflow-idea-entry.mjs --rough-idea "<ROUGH_IDEA>" --visual-target ios_game --execute --json ``` Use a here-doc for multi-line ideas: ```bash node scripts/workflow-idea-entry.mjs --rough-idea-file /tmp/rough-idea.txt --execute --json ``` `--execute` creates only a local sandbox session and the initial `capture` state. It creates **no** Multica Project, parent issue, Controller Autopilot, or Work Agent Spine binding. ## Stage-advance contract - Entry advances the local lane through validated stages without waiting for a per-stage request. - Use `node scripts/workflow-idea-stage-advance.mjs --canary-path <session-path> --to-promotion-ready`. - Do not invoke the legacy `workflow-sandbox-canary.mjs` for a product idea. It is a separate Multica rehearsal harness. - At `build_handoff`, reuse an exact-title planned Multica Project when one exists; otherwise create an implementation Project. Only then bind the implementation work to Spine. ## Bootstrap (plan only / no Multica mutations) ```bash node scripts/workflow-idea-entry.mjs --rough-idea "<ROUGH_IDEA>" --dry-run --json ``` ## What happens before Multica After `--execute`: 1. A local sandbox path and session manifest are created. 2. The session advances through local artifact contracts to `promotion_ready`. 3. No Project, parent issue, Autopilot, resource, or Spine state is created. At `build_handoff`, the implementation lane creates or reuses its Project, attaches implementation resources, and begins using Spine. Until then, the user does **not** need to run `multica` commands. ## Visual and iOS build lane When `--visual-target visual|game|ios_game` is explicit, Idea-to-Build inserts and requires `ui_design_brief` between `design_doc` and `implementation_spec`. The brief must carry: - a persisted design system: semantic palette/surface roles, typography scale, spacing, and anti-patterns; - screen/state matrix and interaction hierarchy; - accessibility rules: contrast, Dynamic Type, VoiceOver order/labels, 44pt controls, color-not-only semantics, and Reduce Motion; - implementation boundaries: preserve mechanics and asset IDs; fail closed when required assets/capabilities are missing; - an evidence plan: named states, device/OS, screenshot/video requirements, commit SHA, and SHA-256 manifest. For iOS work, the implementation agent follows the Tiny Loop Factory reference loop: 1. **UI/UX Pro Max** — generate/persist the design system before screen changes. 2. **iOS UI Debug Loop** — capture symptom → select lane → smallest useful fix → recapture. 3. **iOS Local Build Loop** — run the exact macOS `xcodebuild` command and record pass/fail evidence. 4. **iOS Debugger Agent** — use the booted Mac Simulator for UI tree, screenshots, logs, and gameplay video when runtime evidence is needed. 5. **SwiftUI Expert** — review view structure, Dynamic Type, accessibility order, scroll safety, and API availability. Functional PASS is not Visual PASS. `visual_review` remains a separate target-conditional gate after implementation and verification. The UI brief cannot be skipped, and each local stage advances only after explicit human approval. Missing macOS/Xcode or Simulator evidence blocks the visual lane; no screenshots or findings may be fabricated. Human-owned actions remain merge, signing, TestFlight, App Store, production, billing, secrets, and destructive operations. See [`docs/idea-to-build-visual-ui-workflow.md`](../../docs/idea-to-build-visual-ui-workflow.md) for the complete stage/evidence contract. ## Report back Parse the JSON and tell the user: | Field | Meaning | |-------|---------| | `workflowRunId` | Local idea session id | | `campaign.currentStageId` | Current Hermes stage | | `campaign.workflowStatus` | Run status | Track status any time with: ```bash node scripts/workflow-idea-status.mjs --json ``` Or invoke `/skill:idea-status`. ## Policy - **Never** set `productionAllowed=true`. - Do not rotate secrets, change billing, or run destructive production ops. - This skill targets the local Idea-to-Build lane. The sandbox canary and Maintenance production-run remain separate rehearsal lanes. ## Troubleshooting | Issue | Action | |-------|--------| | `dist/lib` missing | `npm run build` | | prior session already exists | omit `--reuse-default-canary`; each idea gets a fresh session path by default | | wrong session resumed | use `--canary-path` from dry-run JSON | | implementation must begin | complete `build_handoff`; then create or reuse the implementation Project and bind Spine | See [`docs/workflow-idea-entry-live-execute-runbook.md`](../../docs/workflow-idea-entry-live-execute-runbook.md).
From eiei114/pi-multica-spine, MIT.
More from eiei114/pi-multica-spine
Signals
- Stars
- 4
- Forks
- 1
- Open issues
- 1
- Last commit (this folder)
- Jul 26, 2026 (3 months ago)
- Last release
- Oct 20, 2026 (10 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/eiei114/pi-multica-spine
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from eiei114
- 0.609score
maintenance-build
eiei114
Start the Hermes Maintenance Multica workflow after the human explicitly invokes this skill.
SkillAFreshness: A★ 4Claude Code +9 - 0.603score
idea-status
eiei114
Read-only Operations Cockpit for Idea Sessions.
SkillAFreshness: A★ 4Claude Code +9 - 0.589score
template-skill-bridge
eiei114
Template skill contributed by extensions/skill-bridge via resources_discover.
SkillAFreshness: A★ 2Claude Code +9 - 0.581score
spotify-playback
eiei114
Control Spotify playback and share now playing to X from Pi.
SkillAFreshness: A★ 1Claude Code +9 - 0.561score
example-skill
eiei114
Example Agent Skill shipped by this Pi package template.
SkillAFreshness: A★ 2Claude Code +9 - 0.404score
java-junit
eiei114
Get best practices for JUnit 5 unit testing, including data-driven tests
SkillDFreshness: DClaude Code +9
Cite this page
idea-to-build skill. Agent Mart. https://agentmart.dev/skills/eiei114-idea-to-build (last updated 2026-10-24)