Stores everything about the user's machine. Populated once during onboarding. Prevents Claude from ever asking the same environment question across sessions.
{
"checked_at": "2026-03-27",
"os": "linux",
"distro": "ubuntu",
"shell": "bash",
"locale": "fr_FR",
"python": {
"installed": true,
"version": "3.12.3",
"path": "/usr/bin/python3"
},
"vscode": {
"installed": true,
"path": "/usr/bin/code",
"extensions": {
"python": true,
"python_indent": true
}
},
"learning_dir": "/home/charles/learning",
"pip": {
"installed": true,
"path": "/usr/bin/pip3"
}
}{
"name": "Charles",
"started": "2026-06-01",
"experience": "never",
"goal_track": "games",
"session_length": "30min",
"level": 2,
"xp": 340,
"total_sessions": 12,
"total_time_minutes": 380
}{
"current_module": 2,
"current_lesson": 8,
"current_exercise_status": "in_progress",
"completed_lessons": [1, 2, 3, 4, 5, 6, 7],
"completed_projects": ["p1_mad_libs"],
"skipped_lessons": [],
"hint_counts": {
"8": 1
}
}{
"print": {"score": 95, "last_practiced": "2026-06-10", "times_practiced": 8},
"variables": {"score": 90, "last_practiced": "2026-06-10", "times_practiced": 6},
"strings": {"score": 85, "last_practiced": "2026-06-09", "times_practiced": 5},
"if": {"score": 60, "last_practiced": "2026-06-12", "times_practiced": 3},
"elif": {"score": 35, "last_practiced": "2026-06-12", "times_practiced": 1}
}Score update rules:
| Event | Delta |
|---|---|
| First attempt pass (all tests) | +30 |
| Pass after 1 hint | +20 |
| Pass after 2+ hints | +10 |
| Fail 3+ times | -10 |
| Used in project | +15 |
| Correct in spaced review | +5 |
| Forgotten in review | -15 |
| Weekly time decay (per week unused) | -5 |
{
"patterns": [
{
"type": "SyntaxError",
"subtype": "missing_colon",
"occurrences": 4,
"lessons": [7, 8, 9, 11],
"last_seen": "2026-06-12",
"improving": true
},
{
"type": "TypeError",
"subtype": "string_int_concat",
"occurrences": 2,
"lessons": [5, 7],
"last_seen": "2026-06-08",
"improving": false
}
]
}When an error type reaches 3+ occurrences and improving: false, trigger a targeted micro-drill.
{
"terms": {
"variable": {"introduced_lesson": 2, "confidence": 90, "definition": "A name that stores a value"},
"string": {"introduced_lesson": 3, "confidence": 85, "definition": "Text data inside quotes"},
"boolean": {"introduced_lesson": 6, "confidence": 40, "definition": "True or False value"}
}
}{
"current_streak": 5,
"longest_streak": 7,
"last_session_date": "2026-06-12",
"session_dates": ["2026-06-08", "2026-06-09", "2026-06-10", "2026-06-11", "2026-06-12"]
}{
"unlocked": [
{"id": "first_program", "name": "First Program", "date": "2026-06-01"},
{"id": "variable_master", "name": "Variable Master", "date": "2026-06-05"},
{"id": "streak_3", "name": "Streak Starter", "date": "2026-06-10"}
]
}The most granular tracking — every /learn check creates a record. This is how we track the full learning journey within each exercise.
{
"lesson": "07_if_else",
"started": "2026-06-05T14:30:00",
"completed": "2026-06-05T14:52:00",
"total_attempts": 3,
"hints_used": 1,
"attempts": [
{
"attempt": 1,
"timestamp": "2026-06-05T14:35:00",
"code_snapshot": "def can_vote(age):\n if age > 18:\n return 'You can vote!'\n",
"result": "test_failure",
"tests_passed": 3,
"tests_total": 7,
"failing_tests": ["test_exactly_18", "test_years_remaining"],
"error_type": null,
"error_message": null,
"error_line": null,
"hints_used_before": 0
},
{
"attempt": 2,
"timestamp": "2026-06-05T14:41:00",
"code_snapshot": "def can_vote(age):\n if age >= 18:\n return 'You can vote!'\n else:\n return 'Not yet! ' + 18 - age + ' more years.'\n",
"result": "runtime_error",
"error_type": "TypeError",
"error_message": "can only concatenate str (not \"int\") to str",
"error_line": 5,
"tests_passed": 0,
"tests_total": 7,
"failing_tests": [],
"hints_used_before": 1
},
{
"attempt": 3,
"timestamp": "2026-06-05T14:52:00",
"code_snapshot": "def can_vote(age):\n if age >= 18:\n return 'You can vote!'\n else:\n return f'Not yet! {18 - age} more years to go.'\n",
"result": "pass",
"tests_passed": 7,
"tests_total": 7,
"failing_tests": [],
"error_type": null,
"error_message": null,
"error_line": null,
"hints_used_before": 1
}
]
}This example tells the full story:
- Used
>instead of>=(off-by-one) → 3/7 tests pass - Fixed the comparison but hit TypeError building the string → runtime crash
- Switched to f-string → all 7 pass
What we can derive from attempt data:
- Average attempts per exercise (difficulty indicator)
- Common error patterns across exercises
- Improvement trend (fewer attempts over time)
- Hint dependency (using fewer hints as skills build)
- Time-to-completion per exercise
- Which exercises were hardest for this specific learner
{
"date": "2026-06-12",
"duration_minutes": 35,
"lessons_attempted": [8],
"lessons_completed": [],
"concepts_practiced": ["elif"],
"errors_encountered": [
{"type": "SyntaxError", "subtype": "missing_colon", "line": 5}
],
"hints_used": 1,
"xp_earned": 0,
"notes": "Working on elif chains. Got stuck on grade boundaries."
}| Actor | Can Read | Can Write |
|---|---|---|
| Claude | Everything | .claude/ state files, new exercise/test files |
| User | Everything | Their own exercise .py files ONLY |
| Test runner | Exercise files | stdout only |
Critical: Claude NEVER modifies an exercise file that the user has started editing. Claude only creates new scaffold files and updates state.