Execute qualquer Skill no Manus
com um clique

Execute qualquer Skill no Manus com um clique

caspar-prototype

Estrelas0

Forks0

Atualizado3 de junho de 2026 às 17:16

Generate a single self-contained, clickable HTML prototype to make a feature visible before planning — resolving ambiguity pre-scope, validating flows pre-plan, or rendering an approved UX spec for stakeholder review. Triggers on "prototype this", "mock up the UI", "make a clickable preview", or a mid-flow `ux` Stage 1→2 transition. Do NOT trigger to write production app code, build a multi-file frontend, or plan/scope the work itself — this produces one throwaway HTML file only.

Instalação

Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.

Executar no Manus

Fonte

Codename-Inc

Codename-Inc/caspar

Abrir repositório GitHub Ver repositórios do creator

Download

Executar no Manus

Ocupações relacionadasSOC

Baseado na classificação ocupacional SOC

Desenvolvedores webInformática e Matemática·SOC 15-1254

SKILL.md

readonly

Mais deste repositório

mesmo repositório

feature-codex-caspar-implementation

Codename-Inc/caspar

Use when modifying the Codex CASPAR install flow, project skill syncing, learn/recall registry files, or Codex-specific runtime files. TRIGGER when: codex, caspar, codex install, project skills, registry, caspar-learn, caspar-recall, config.toml, doctor

2026-07-100

caspar-code-review

Codename-Inc/caspar

Run an independent adversarial code review of completed work using a pinned high-effort opposing runtime, with a same-contract native fallback. Finds bugs, correctness and security failures, regressions, performance landmines, overengineering, and missing behavioral tests. Review only; never edits code.

2026-07-100

caspar-execute

Codename-Inc/caspar

Build the planned feature from a compact execute.md index and sliceable tasks.json detail, dispatching parallel task waves through deterministic checks, lightweight sentinel review, and final adversarial review/validation. Trigger after execute.md/tasks.json exist or to resume a partially-built wave plan. Do NOT trigger for scoping/planning, unplanned bug-fixing, or dead-code cleanup (caspar-prune).

2026-07-100

caspar-guide

Codename-Inc/caspar

Use when rendering the Next Steps footer at the end of any caspar command, suggesting which caspar command to run next, or answering "what command should I run?" / "how does the workflow work?". Do NOT use mid-task or for executing a phase — this only renders guidance and the footer.

2026-07-100

caspar-plan-review

Codename-Inc/caspar

Independent adversarial review of plan.md before task generation. Use a pinned high-effort opposite-runtime reviewer; fall back to one native reviewer with the same contract when unavailable. Trigger after create_plan and before create_tasks. Do NOT review generated tasks, finished code, or apply scope changes.

2026-07-100

caspar-task-review

Codename-Inc/caspar

Adversarial review of generated execute.md + tasks.json against a reviewed plan using a pinned high-effort opposite runtime, with a same-contract native fallback. Checks translation, criteria, RED pairing, dependencies, waves, and alignment. Do NOT review plan quality, finished code, or change scope.

2026-07-100

name	caspar-prototype
description	Generate a single self-contained, clickable HTML prototype to make a feature visible before planning — resolving ambiguity pre-scope, validating flows pre-plan, or rendering an approved UX spec for stakeholder review. Triggers on "prototype this", "mock up the UI", "make a clickable preview", or a mid-flow `ux` Stage 1→2 transition. Do NOT trigger to write production app code, build a multi-file frontend, or plan/scope the work itself — this produces one throwaway HTML file only.
user-invocable	true

prototype

Produce one portable, self-contained HTML file that makes a feature visible before it is planned. Output saved to {OUT_DIR}/prototypes/{slug}_{MMDDYY}.html.

Inputs

$ARGUMENTS (feature description, --explore flag, optional output path).
Env markers: FROM_UX=true, FROM_KICKOFF=true.
On-disk context (read FULLY, no offset/limit): docs/tasks/{branch}/concepts/scope.md, .../specs/prd.md, .../ux.md. Resolve branch via git rev-parse --abbrev-ref HEAD at runtime.

Mode + fidelity (detect most-context-first; this drives intake and faithfulness)

Mode	Signal	Fidelity	Job
post-ux	complete `ux.md` (has Screens, Layouts, Components, Interactions, States, Content)	high-fi	render the spec faithfully — do not invent screens or contradict documented states/copy
flows-only ux	`FROM_UX=true`, or ux.md with flows but missing Stage-2 sections	mid-fi	flows are approved input; you make the UI calls
explore	`--explore` flag	low-fi (grayscale, layout-only)	concept unvalidated
post-scope	scope.md exists, no ux.md	mid-fi	validate scope
standalone	none	mid-fi	ask what we're prototyping

Working set

Stage 1 intake (tight: 2–4 questions, never a form), then a gated parallel research+generate pass. Single-threaded write of the HTML.

Method / guardrails

Immediate reply, then detect. Respond before any tool call (except reading ux.md when FROM_UX=true). If ARGUMENTS empty → ask what we're prototyping.
Read context before asking. Read whichever of scope.md/prd.md/ux.md exist, fully; classify ux.md as complete / flows-only / absent → sets mode.
Confirm fidelity + visual anchor. Recommend fidelity per the table; request a visual anchor (brand colors, font stack, reference URL, or named aesthetic e.g. "Linear-style"). If user skips, commit to a deliberate named aesthetic and call it out at the top of the file. Gate: wait for confirmation (auto-proceed only when FROM_UX=true).
Dispatch parallel subagents (roster fixed — do not add or rename):
- @web-research — 2–3 living UI references + the established convention for this interaction type + concrete palette/type/layout decisions for the anchor (hex, font families). Anti-slop forcing function: without it the model defaults to Inter + purple gradients. <400 words, cite sources.
- @analyst — extract mode (post-ux): pull Screens/Layouts/Components/Interactions/States/Content verbatim from ux.md; use documented copy EXACTLY; flag spec-silent details as "filled assumptions"; never invent screens. synthesize mode (else): primary flow, per-screen state list, realistic domain content (no Lorem ipsum), component inventory.
- @patterns — only if an existing app to anchor to: real design tokens, component conventions, interaction patterns (return actual hex/class values, not paths). Skip otherwise. Return findings in-thread (compressed); write no intermediate docs.
Generate after ALL subagents return. Use @dev (or inline if trivial). Embed the negative constraints below verbatim in the dev prompt.

Required HTML structure (load-bearing — these prevent the known failure modes)

<!DOCTYPE html> + minimal <head> (<meta viewport>, inline <style>).
Metadata comment at top of <head>: Feature · Fidelity · Generated (date) · Branch · Flow covered · Screens/states · Visual anchor · Source spec (ux.md path or "synthesized from …") · Key assumptions · Filled assumptions (post-ux only) · NOT included · Next step.
Design-tokens comment (Primary/Accent/Surface/Text/Font/Border-radius/Spacing) AND the same encoded as CSS custom properties on :root.
Nav bar only if multi-screen — vanilla-JS display:block/none toggling. No router, no framework.
One <section> per screen, in user-encounter order. Every screen shows the happy path AND ≥1 of: empty / error / loading state.
Inline <script> at end of body for interactions. Vanilla JS only.

Hard constraints (positive form unless destructive)

Single self-contained file, under 300KB. Tailwind-via-CDN OK for mid/high-fi; skip for low-fi.
Portable assets only: inline SVG / data-URI / CSS shapes. No <img src="http…">, no local relative paths (./, ../), no custom WOFF. Fonts: one Google-Fonts family (≤2 weights) or system.
No generic AI aesthetic (no default Inter/Roboto, no purple-on-white gradients). No Lorem ipsum / placeholder filler. No happy-path-only screens. No broken interactivity (no href="#" jumps, no console errors). No inconsistent components — each recurring element = one named CSS class, reused, never re-styled inline.

Outputs + DONE

File at {OUT_DIR}/prototypes/{slug}_{MMDDYY}.html (OUT_DIR=docs/tasks/{branch} unless FROM_UX/FROM_KICKOFF reuse the task dir or user gives a path; mkdir -p the prototypes subdir). DONE when:

Mode + fidelity detected; context docs read fully before any question; visual anchor captured (never left generic).
Subagents dispatched in parallel and all completed before generation.
HTML is one self-contained file <300KB; metadata + design-token blocks present; tokens on :root.
Every screen: happy path + ≥1 non-happy state; realistic content; inline SVG only; recurring components share one class; all interactions respond, no console errors.
Portability self-check run (size, external images, relative paths, component reuse, state coverage); any violation surfaced as a caveat, not silently shipped.

Handoff

Present: path, size, screen count, fidelity, visual anchor, key assumptions, NOT-included list; tell user to open it in a browser and that it is fully shareable. Iterate on feedback (small tweaks → edit HTML; structural → re-run subagents + regen; re-run the portability check after any edit). post-ux only: surface filled assumptions and offer to promote selected ones into ux.md (edit the matching section, re-present with a one-line diff). Close with an inline Next-Steps line, mode-specific: FROM_UX → resume caspar-ux Stage 2; --explore → caspar-scope; post-scope → caspar-ux or caspar-plan.

Escalate if

Spec contradicts the prototype (post-ux): ask which is authoritative per contradiction, then update ux.md and the HTML together and re-run the portability check.
No usable visual anchor and no inferable aesthetic, or required context docs are missing/unreadable when a mode expects them.