david-sql-task-importance

name	david-sql-task-importance
description	Use when David wants to score or rate the importance of his Todoist tasks in docs/data/todoist-task-dimensions.db. Presents tasks 7 at a time (A-G), David gives scores (0-10), AI saves them. Also updates brain/ and docs/data/user-behavior.db.

David SQL Task Importance — Batch Scoring

Goal: Score every task in docs/data/todoist-task-dimensions.db with a david_importance value (0-10 integer). David scores, AI records. No guessing.

Why This Matters

david_importance is fundamentally different from daily priority lists. Priority = "what should I do today?" Importance = "how much does this matter to my life, goals, and business?" A task can be low-priority today but importance-10. This column captures what David truly cares about — the signal that no daily triage can surface. Tasks scored 0 are archived to not_relevant automatically.

Scale

Score	Meaning
0	Dead. Irrelevant. Completed with no future bearing.
1	Lowest leverage. Taking out trash, vacuuming. Does not matter long-term.
5	Mid-range. Has some value but not a priority mover.
10	Massive implications to David personally, the business, or the long-term vision/mission.

Rules

• AI NEVER guesses scores. Only David assigns david_importance. AI presents, David scores, AI saves.
• Check DB before asking questions. Before presenting any task, check if david_reasoning and other columns are already populated. Never ask redundant questions.
• Voice-to-text decimals are errors. If David says "8.8" or "7.7", assume the repeated digit is a transcription artifact. Always round DOWN (8.8 → 8, 7.7 → 7).
• Voice-to-text duplicates are confirmation. If David repeats a score ("Task E, 9. Task E, 9."), take the first instance.

Presentation Format

Present 7 tasks per batch, lettered A-G:

**X/Y scored, Z remaining**

A. task name here
B. task name here
C. task name here
D. task name here
E. task name here
F. task name here
G. task name here

• Always show progress (scored / total, remaining) before each batch.
• Letters A-G only. No numbers. David uses voice-to-text and letters are clearer.
• Task name only. No project, no priority, no metadata. Keep it minimal.

Scoring Method (CRITICAL — NO OFFSETS)

Never use SQL OFFSET to map scores to tasks. Offsets shift when rows are added/deleted and cause cascading errors.

Step 1: Fetch with explicit IDs

SELECT DISTINCT todoist_id, task_name
FROM tasks
WHERE todoist_id IS NOT NULL AND david_importance IS NULL
GROUP BY todoist_id
ORDER BY id
LIMIT 7;

Store the 7 todoist_id values explicitly before presenting to David.

Step 2: Present task names (A-G)

Step 3: David gives scores

Step 4: Update using explicit todoist_id

Write each UPDATE as a separate, explicit SQL statement:

UPDATE tasks SET david_importance = 8 WHERE todoist_id = '<TODOIST_ID_1>';
UPDATE tasks SET david_importance = 0 WHERE todoist_id = '<TODOIST_ID_2>';

Never use subqueries, offsets, or shell loops. One todoist_id per UPDATE. No ambiguity.

Step 5: Auto-advance

After saving, immediately present the next 7. No confirmation needed.

Source of Tasks

Default: all tasks in docs/data/todoist-task-dimensions.db where david_importance IS NULL.

David may specify a different scope:

• "score all inbox tasks" → filter by domain or project
• "re-score my 9s and 10s" → filter by existing score
• "score today's Todoist tasks" → fetch fresh from Todoist API first (use todoist skill), upsert into DB, then score

Fetching Fresh Tasks from Todoist

When David wants to score tasks not yet in the DB:

1. Fetch from Todoist API using the todoist skill
2. Check each task against DB by todoist_id
3. INSERT new tasks (with whatever columns can be inferred from task text)
4. Present for scoring

Companion Updates

After each batch of David's responses:

• docs/data/user-behavior.db — log a message entry if behavior-tracking is active
• brain/ — update relevant files if David's scores reveal new patterns about how he thinks or prioritizes

Database Location

docs/data/todoist-task-dimensions.db — column: david_importance INTEGER DEFAULT NULL

Wrap Up

When all tasks scored:

• Report total scored, distribution (how many 0s, how many 10s, etc.)
• Show top 20% (scores 8-10) as a ranked list
• Stage and push to GitHub