Progress

CRITICAL RULES

• READ-ONLY: Never write to or modify any file. No JSON updates, no MD regeneration, no session tracking.
• No conversation analysis: Do not scan user messages for language errors. This is purely historical analytics.
• No coaching blocks: Do not append correction, teaching, immersion, or SRS review blocks.
• Data as-is: Report the data from the JSON files exactly. Do not infer corrections or create new patterns.

Flag Mapping

• en → 🇬🇧 English
• es → 🇪🇸 Español
• fr → 🇫🇷 Français
• de → 🇩🇪 Deutsch
• it → 🇮🇹 Italiano
• ja → 🇯🇵 日本語
• ko → 🇰🇷 한국어
• nl → 🇳🇱 Nederlands

Report Format

Generate the following sections in order. Omit a section only when noted.

Section 1: Header

{flag} {Language} — Progress Report
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Level: {level} | Active since: {active_since} | Sessions: {stats.total_sessions}

Section 2: Overview (always shown)

## Overview
- Learning {language} since {active_since} ({N days/weeks/months ago})
- {stats.total_sessions} coaching sessions across {count unique projects in sessions} project(s)
- {stats.total_corrections} corrections | {stats.patterns_active} active patterns | {stats.patterns_resolved} resolved
- {stats.vocabulary_size} vocabulary terms acquired

Calculate time elapsed from active_since to today. Use "today" if same day, "N days" if < 7, "N weeks" if < 30, "N months" otherwise.

Count unique projects from sessions[].project values (excluding duplicates).

Section 3: Pattern Analysis (shown when patterns array is non-empty)

## Pattern Analysis

### Active Patterns ({count where resolved=false})
| Pattern | Type | Corrections | First Seen | Last Seen | SRS Status |
|---------|------|-------------|------------|-----------|------------|

Sort by times_corrected descending (most persistent first). For SRS Status column:

• If next_review is not null: show "Due: {date} ({interval_days}d, ease {ease_factor})"
• If next_review is null: show "—"

### By Type

Group patterns by type. For each type: count patterns and sum times_corrected. Show as:

• {type}: {count} pattern(s) ({sum} corrections) — {percentage}% of total

Sort by sum descending. Calculate percentage as sum / stats.total_corrections * 100.

### Most Persistent

Show top 3 patterns by times_corrected (skip if fewer than 1):

1. **{native_form}** — {times_corrected} corrections, {times_correct_since_last_error} correct usages
   {explanation}

### Resolved Patterns ({count where resolved=true})

If resolved > 0: show table with native_form, target_correction, times_corrected, last_correct_usage (as resolution date).

If resolved == 0: show "No patterns resolved yet. Patterns resolve after 21+ days of correct usage with 5+ consecutive correct uses."

Section 4: SRS Status (shown when any pattern has non-null next_review)

## SRS (Spaced Repetition)

### Due for Review

List patterns where next_review <= today and resolved=false, sorted by next_review ascending:

- **{native_form}** — {overdue_days} day(s) overdue (due: {next_review}, interval: {interval_days}d, ease: {ease_factor})

If next_review == today: show "due today" instead of "0 day(s) overdue".

If none due: show "No reviews due today."

### Upcoming Reviews

List patterns where next_review > today and resolved=false, sorted by next_review ascending:

- **{native_form}** — due {next_review} (interval: {interval_days}d)

If none upcoming: show "No upcoming reviews scheduled."

### SRS Health
- Patterns under SRS: {count with non-null next_review and resolved=false}
- Average ease factor: {mean of ease_factor, rounded to 2 decimals}
- Interval range: {min interval_days}d – {max interval_days}d

If no patterns have SRS fields activated: show "SRS will activate after your first pattern correction." instead of the whole section.

Section 5: Vocabulary (shown when vocabulary array is non-empty)

## Vocabulary ({vocabulary array length} terms)

| Term | POS | Translation | Taught | Shown | Used |
|------|-----|-------------|--------|-------|------|

For each vocabulary entry, extract fields with fallbacks for legacy schema:

• Term: target_term or term
• POS: part_of_speech or pos
• Translation: source_term or translation_en or translation_ptbr
• Taught: first_taught (short date format)
• Shown: times_shown
• Used: times_used_by_user or 0

Sort by first_taught descending (newest first).

### Acquisition Rate
- {vocabulary_size / total_sessions} terms per session (avg)

If total_sessions is 0, show "—" instead of dividing by zero.

If vocabulary array is empty: show "No vocabulary taught yet." and explain based on mode:

• If mode is corrective: "Your {language} mode is corrective — only mistakes are corrected. Switch to both for vocabulary teaching."
• If mode is active or both: "Vocabulary will be acquired through active teaching and immersion blocks over time."

Section 6: Session Timeline (shown when sessions array is non-empty)

## Session Timeline

### Recent Sessions
| Date | Project | Corrections | Vocab Taught | SRS Reviews | Notes |
|------|---------|-------------|-------------|-------------|-------|

Show the last 10 sessions from the sessions array (sorted by date descending). For each session:

• Corrections: length of patterns_addressed
• Vocab Taught: length of vocabulary_taught
• SRS Reviews: srs_reviews
• Notes: truncate notes to 40 chars if longer

If more than 10 sessions: add "(showing last 10 of {N} sessions)" below the table.

### Activity
- Sessions this week: {count where date is within current ISO week}
- Sessions this month: {count where date is within current month}
- Current streak: {consecutive calendar days with sessions, counting back from most recent} day(s)
- Longest streak: {max consecutive calendar days with sessions} day(s)

For streak calculation: check consecutive dates in the sessions array (each date counts once regardless of multiple entries). Work backward from the most recent session date. If only 1 session exists, streak is "1 day".

If current streak equals longest streak, omit "Longest streak" line.

Section 7: Insights (always shown)

## Insights

Generate 3-5 actionable insights based on the data. Choose from these templates (in priority order):

1. Weakest pattern type (when patterns exist): Identify the type with the highest total corrections. "Weakest area: {type} ({sum} of {total} corrections, {pct}%). {brief advice}."

2. Native interference (when patterns of type interference, false_friend, or spelling with L1 influence exist): "Native interference: {count} patterns show direct {native_language} influence. {brief explanation of why this is normal for their level}."

3. SRS trajectory (when SRS patterns exist): Comment on SRS state — all new (interval=1d), progressing (mixed intervals), or nearing resolution (interval >= 14d).

4. Vocabulary-correction balance (when one side is 0): If vocabulary is 0 but patterns > 0, suggest mode change. If patterns are 0 but vocabulary > 0, celebrate clean usage.

5. Session frequency (when sessions > 3): Comment on regularity. Daily sessions accelerate SRS. Gaps reset momentum.

6. Resolution progress (when any pattern has times_correct_since_last_error >= 3): Highlight patterns close to resolving.

7. Early stage (when total_sessions < 5): "Getting started: {N} sessions so far. Patterns build naturally with daily use. First resolutions typically happen after 3-4 weeks."

8. Immersion active (when the language config has immersion: phrase or immersion: sentence): Note that immersion mode is active and contributing to vocabulary acquisition.

Rules:

• Maximum 5 insights per language
• Be encouraging about progress, honest about persistent patterns
• Use data to support each insight (cite numbers)
• Do not repeat information already visible in the tables above — add interpretation

Close the report with:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Cross-Language Summary (for all mode only)

After all individual language reports, add:

## Cross-Language Summary
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

| Language | Level | Sessions | Patterns (active/resolved) | Vocabulary | Last Active |
|----------|-------|----------|---------------------------|------------|-------------|

One row per configured language. Last Active = stats.last_session or "never" if null.

Important

• This is a PURE ANALYTICS skill — no corrections, no teaching, no file writes
• Focus on longitudinal trends, not single-session observations
• All dates displayed in short format (e.g., "Feb 22") for readability
• When data is sparse, acknowledge it honestly and set expectations
• Use tables for structured data, prose for insights