Skip to content
All Skills

Fluent Writing

Run an interactive writing practice session (emails, letters, forms, short texts) with systematic error analysis, category-tagged corrections, and detailed feedback. Triggered only when the learner types /fluent-writing. Selects a scenario matched to mastery, lets the learner compose, then analyzes grammar, register, vocabulary, structure, and spelling before updating all databases.

Education & Writing|v1|Updated 7/14/2026|GitHub source
MCP get_skill({ skillId: "fluent-writing-512b32fb" })

Use this skill with your agent

Create a free account and connect via MCP

Get Started Free
# Writing Practice Session

## Overview

Full-text writing practice with systematic correction. One scenario per session, detailed feedback broken down by severity and category, DB update at end. Mastery-driven scenario selection keeps the task at the right level — challenging, not frustrating.

## When to Use

Trigger this skill only when the learner types `/fluent-writing`. The skill is gated with `disable-model-invocation: true` — a 15-20 min interactive session with DB writes should never start from an ambiguous prompt.

Skip this skill in favor of `/fluent-vocab` if the learner has not yet hit mastery 2 in basic vocabulary — writing needs a minimum word bank.

## Instructions

### 1. Load context

```bash
python3 "${CLAUDE_PLUGIN_ROOT:-${CLAUDE_PROJECT_DIR:-.}}/.claude/hooks/read-db.py"
```

Need: `learner-profile` (level, target language, focus areas), `mistakes-db` (weak writing patterns), `mastery-db` (writing sub-skills).

### 2. Pick scenario type

From `mastery-db.skills_mastery`:

- Formal email (if `writing_formal_email` mastery < 4)
- Informal email (if `writing_informal_email` < 4)
- Form filling (if `writing_forms` < 4)
- Newsletter / personal text (if overall writing < 3)
- Mixed scenarios (if all ≥ 4)

Scenarios must match the learner's CEFR level — A2 uses everyday situations, B1+ adds opinion / complaint / inquiry.

### 3. Present the task

```markdown
## ✍️ Writing Exercise

**Scenario:** {clear description in native language}

**Task:** Write a {type} in {target_language}.

**Requirements:**
- Length: {X-Y} words
- Include: {must-include elements}
- Register: {formal / informal}
- Level: {CEFR}

{Optional: example structure for harder tasks}

**Write your {text_type} below:**
```

### 4. Wait for the full text

Don't correct mid-composition. Let the learner finish.

### 5. Systematic error analysis

Check every sentence for these categories:

1. **Grammar** — word order, conjugation, clause structure, articles
2. **Formal/informal** — register consistency
3. **Vocabulary** — wrong word, English mixing, register-wrong synonyms
4. **Missing elements** — greeting, closing, required fields
5. **Spelling** — minor at A2, weightier at B2+
6. **Structure** — organization, flow, paragraphing

Tag each finding with a severity: 🔴 critical, 🟡 moderate, 🟢 minor.

### 6. Detailed feedback

Diverges slightly from the standard `fluent-feedback-formatter` template because writing answers are multi-sentence. Use this variant:

```markdown
## Feedback

### ✅ What You Did Well
- {strength 1}
- {strength 2}

### ❌ Areas to Improve

**Critical:** 🔴
- {issue}: "{wrong}" → **"{correct}"** — {why}

**Moderate:** 🟡
- {issue}: {explanation}

**Minor:** 🟢
- {spelling / punctuation}

### 📝 Corrected Version

```
{fully corrected text}
```

**Score: {X}/10**

**Breakdown:**
- Grammar: {Y}/10
- Vocabulary: {Z}/10
- Structure: {W}/10
- Communication: {V}/10

---
```

### 7. Optional rewrite

If score < 7, offer:

```markdown
**Want to try again?** Rewriting with the corrections locks in the patterns.

Type "rewrite" to try again, or "next" to continue.
```

### 8. Session summary

```markdown
## 📊 Writing Session Summary

**Text Type:** {type}
**Score:** {X}/10
**Key Takeaways:**
- {learning 1}
- {learning 2}
- {learning 3}

**Next Time:**
- Focus on: {weak pattern}
- Review: {relevant flashcards}

{target-language "well done"}! ✍️
```

### 9. Update all databases

Use the `fluent-db-updater` skill:

- `command_used: "/fluent-writing"`, `skills_practiced: ["writing"]`
- `skill_scores.writing: {exercises: 1, correct: 1_if_score_≥_7_else_0, time_minutes}`
- `errors[]` — one per distinct pattern found (dedupe; the script bumps frequency)
- `focus_next_session[]` — top 2 patterns to drill

Also save the exchange as `/results/fluent-writing-session-{NNN}.md` with the full task, the learner's original text, the corrected version, and the error table. The `fluent-session-analyzer` skill depends on this format.

## Examples

### Example 1 — formal email task (Dutch A2)

> ## ✍️ Writing Exercise
>
> **Scenario:** You need to cancel a doctor's appointment you have on Thursday.
>
> **Task:** Write a formal email to the doctor's office in Dutch.
>
> **Requirements:**
> - Length: 40-60 words
> - Include: greeting, reason, request for new appointment, closing
> - Register: formal
> - Level: A2
>
> **Write your email below:**

Learner: "Hallo, Ik schrijf je omdat ik kan niet komen op donderdag. Ik ben ziek. Kunnen we een andere dag afspreken? Bedankt, Mohammad"

> ## Feedback
>
> ### ✅ What You Did Well
> - Clear reason for cancellation.
> - Correct request structure.
>
> ### ❌ Areas to Improve
>
> **Critical:** 🔴
> - `formal_informal`: "Hallo" + "je" — formal email needs "Beste meneer/mevrouw" + "u". Mixed registers cost exam points.
> - `grammar`: "ik kan niet komen" is OK standalone, but after "omdat" Dutch word order flips: **"omdat ik niet kan komen"**.
>
> **Moderate:** 🟡
> - `missing`: no formal closing. Use "Met vriendelijke groet,".
>
> **Minor:** 🟢
> - Capitalize "Donderdag"? Actually no — Dutch doesn't capitalize weekdays. You were right.
>
> ### 📝 Corrected Version
>
> ```
> Beste meneer/mevrouw,
>
> Ik schrijf u omdat ik niet kan komen op donderdag. Ik ben ziek.
> Kunnen we een andere dag afspreken?
>
> Met vriendelijke groet,
> Mohammad
> ```
>
> **Score: 6/10**
> - Grammar: 6/10 — omdat word order tripped you up.
> - Vocabulary: 8/10 — solid word choice.
> - Structure: 5/10 — missing proper opening + closing.
> - Communication: 7/10 — message was clear despite issues.

## Critical Rules

- **One scenario per session.** Don't chain multiple writing tasks — depth over breadth.
- **Wait for the full answer** before correcting.
- **Severity tagging is mandatory.** Fed into `mistakes-db` and drives spaced repetition priority.
- **Always save the session file** in `/results/` for later analysis by `fluent-session-analyzer`.
- **Never auto-invoke.** This skill is gated; must fire only on explicit `/fluent-writing`.

## Language Reference

### Dutch A2 patterns

**Formal emails:** always "u" (not "je"); open `Beste meneer/mevrouw {NAME},`; closing `Met vriendelijke groet,` + name.

**Informal emails:** "je" not "u"; open `Hallo {NAME},`; closing `Groetjes,` or `Tot snel,`.

**Common mistakes:** mixing formal/informal in one text; word order in `omdat` clauses (verb last); time expressions (`om 10:00 uur`, `op dinsdag`).

Add similar sections for other target languages as the learner needs them.
#language-learning#vocabulary#flashcards#speaking#writing#spaced-repetition#language#learningpythonbash

Related Skills

More skills in Education & Writing

Academic Plotting

Generates publication-quality figures for ML papers from research context. Given a paper section or description, extracts system components and relationships to generate architecture diagrams via Gemini. Given experiment results or data, auto-selects chart type and generates data-driven figures via matplotlib/seaborn. Use when creating any figure for a conference paper.

#broad-capability#ai-researchMIT

Add Educational Comments

Add educational comments to the file specified, or prompt asking for file to comment if one is not provided.

#github-copilot#educationMIT

AI Prompt Engineering & Safety Best Practices

Comprehensive best practices for AI prompt engineering, safety frameworks, bias mitigation, and responsible AI usage for Copilot and LLMs.

#github-copilot#awesome-copilotMIT

Anti Slop

Comprehensive toolkit for detecting and eliminating "AI slop" - generic, low-quality AI-generated patterns in natural language, code, and design. Use when reviewing or improving content quality, preventing generic AI patterns, cleaning up existing content, or enforcing quality standards in writing, code, or design work.

#broad-capability#mathMIT

Apple Notes

Create, view, edit, delete, search, move, or export Apple Notes via the memo CLI on macOS.

#broad-capability#browserMIT

Asc Whats New Writer

Generate engaging, localized App Store release notes (What's New) from git log, bullet points, or free text using canonical metadata under `./metadata`. Optionally pairs with promotional text updates.

#rudrankriyam-app-store-connect#iosMIT