Find agent skills by outcome

- Generate a simple

practical starter prompt

- Lightweight/chat-only models (older 7B–13B models

mini variants) — usually lack depth/context/tool reliability

- Mistral Large 2 / Devstral — Very strong price/performance

fast

- Llama 4 / Llama 405B variants (Meta) — Best open-source frontier performance

excellent for self-hosting

- Gemini 2.5 Pro / Flash (Google) — Outstanding multimodal (especially video/document understanding)

very large context windows

Productivity

- GPT-5 / o3 family (OpenAI) — Deepest reasoning on very complex structured tasks

best at following extremely long/complex instructions

- Grok 3 / Grok 4 (xAI) — Excellent reasoning

real-time knowledge via X

- Claude 4 Opus/Sonnet (Anthropic) — Exceptional long-form reasoning

writing quality

- Tool use

function calling

- Privacy

open behavior

Current Best-in-Class Ranking (January 2026 – general guidance

always tailor to the process):

- Call out misconceptions

dependencies

If AI is viable

recommend which AI engines are best suited and why.

- Track KPIs (e.g.

20% time savings

| Data quality | 3 | Mostly clean

some inconsistency |

Summarize the overall feasibility score (weighted average)

then issue your verdict with clear reasoning.

Interpretation: Clear patterns exist

but context accuracy is critical. Recommend hybrid approach with AI drafts + human review.

| 4 – High Feasibility | Clear logic

consistent data

When scoring

evaluate these dimensions (suggested weights for averaging: e.g.

| 5 – Excellent Feasibility | Predictable process

well-defined data

| 1 – Not Feasible | Process heavily dependent on expert judgment

implicit knowledge

| 2 – Low Feasibility | Some structured elements exist

but goals or data are unclear. AI could assist with insights

Creative

Copilot 規範層（全タスク共通で短く強い指示）

- 仕様の入口は `.github/copilot/00-index.md`。参照順に従い、設計→実装の2段階ループを厳守する。

../claude.md

generic skill