Update worklog files by moving tasks between todo/doing/done states. Use when recording task progress, starting new work, or marking tasks complete. Requires explicit arguments: worklog [done|doing|todo] [description].

Feb 19, 2026

Find agent skills by outcome

Always explain your ranking in the specific context of the user's process

- Generate a simple

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

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

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

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

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

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

- Tool use

- Privacy

- Call out misconceptions

- Track KPIs (e.g.

If AI is viable

**Interpretation:** Clear patterns exist

| Data quality | 3 | Mostly clean

Summarize the overall feasibility score (weighted average)

When scoring

| **5 – Excellent Feasibility** | Predictable process

| **4 – High Feasibility** | Clear logic

| **2 – Low Feasibility** | Some structured elements exist

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

Copilot 規範層（全タスク共通で短く強い指示）

../claude.md

worklog

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

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

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

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

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

Interpretation: Clear patterns exist

| 5 – Excellent Feasibility | Predictable process

| 4 – High Feasibility | Clear logic

| 2 – Low Feasibility | Some structured elements exist

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