メニュー
Slop-prevention middleware. Run before any design or code task to gather sources, then produce a Plan Mode block before touching any file. Eliminates hardcoded colours and generic components caused by agents assuming design decisions.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Route complex requests to the right specialist agent or chain of agents. This skill acts as the central brain of an agent swarm — it analyses what the user needs, determines which specialist domain(s) are required, and coordinates parallel or sequential agent execution. Use this skill when a request spans multiple domains (e.g., "research competitors and create a pitch deck"), when you need to decide which specialist should handle an ambiguous request, or when a task requires a multi-step pipeline across different skills. Triggers on: multi-step requests, cross-domain tasks, "coordinate", "plan this out", "I need help with multiple things", or any complex request that touches more than one specialist area. Also triggers when the user seems unsure which tool or approach to use.
>-
Hybrid DAG execution primitive combining deterministic and agentic nodes with hard iteration caps
">"
Act as a brand ambassador — create authentic, platform-specific social media content that embodies a brand's identity and connects with audiences. Use this skill whenever the user asks to "create social media content for a brand", "act as a brand ambassador", "write ambassador posts", "promote [brand] on social media", "create influencer-style content", "write authentic brand content", "social media ambassador", or any request involving representing a brand through social content. Also triggers on "ambassador voice", "brand promotion posts", "influencer content", "authentic brand posts", "UGC-style content", or when someone wants social media content that sounds like a real person recommending a brand rather than corporate marketing copy. Even if the user just says "help me promote [brand]" or "I need content for [brand]'s social channels" — use this skill.
>-
| name | context-protocol |
| description | Slop-prevention middleware. Run before any design or code task to gather sources, then produce a Plan Mode block before touching any file. Eliminates hardcoded colours and generic components caused by agents assuming design decisions. |
| triggers | ["any /new-feature invocation","any /discuss invocation","any task involving UI, components, or code modification"] |
| context | fork |
Force agents to gather evidence before generating. No file is written until a Plan Mode block is shown and the user approves. This prevents AI SLOP: hardcoded colours, generic components, and assumed design decisions.
Scan the task description for DESIGN triggers:
logo, icon, UI, component, colour, color, style, animation, layout,
design, theme, typography, font, button, card, modal, landing page,
hero, navbar, sidebar, dashboard
Match case-insensitively.
Read apps/web/lib/design-tokens.ts. Extract:
If the file does not exist, note "design-tokens.ts not found" and ask the user where to find the project's colour tokens.
Scan the task description for:
"What's the reference? (URL, image URL, or describe the style)"
Wait for the answer. Do NOT ask a second question.
Read the ## Colour System section of .skills/custom/scientific-luxury/SKILL.md.
Read that section only — not the full file.
If the ## Colour System heading is not found, read the first 60 lines of the file only.
Output this block before writing a single line of code:
CONTEXT GATHERED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✓ Design tokens read: [list key values found — e.g. "bg #050505, cyan #00F5FF, rounded-sm"]
✓ Reference: [URL fetched / image URL / style description — if a style description was provided with no URL, quote it verbatim here]
✓ Colour system: Scientific Luxury — OLED black, spectral accents, Framer Motion only
PROPOSED APPROACH
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[2-4 sentences describing exactly what will be built, using actual token values]
Does this look right? Shall I proceed?
Do NOT write any file until the user confirms. If they change direction, update the proposed approach and show the block again.
From the task description, identify file candidates using these signals:
apps/web/ and apps/backend/src/Read the identified files silently. No output during this step.
"Which files should I look at?"
Wait for the answer. Do NOT ask a second question.
CONTEXT GATHERED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✓ [filename]: [one-line summary of what was found — e.g. "middleware.ts:47 — refresh not awaited"]
✓ [filename]: [one-line summary]
PROPOSED APPROACH
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[2-4 sentences describing exactly what will change and why, referencing actual line numbers]
Does this look right? Shall I proceed?
Do NOT modify any file until the user confirms. If the user redirects, revise the PROPOSED APPROACH section and show the block again.
apps/web/lib/design-tokens.ts