sqlite-provider

name	sqlite-provider
description	SQLite-specific provider implementation for waggle. Loaded when the active provider is sqlite.
user-invocable	false

Waggle — SQLite Provider

This file contains all SQLite-specific implementation details for waggle. Load this file when the active provider is sqlite.

Silent operation: This skill runs as an internal step of an invoking skill. Return results to the invoking flow without user-facing narration — the caller owns all user communication. Only errors, warnings, and prompts required to proceed may surface directly.

Config Retrieval

When detecting-provider requests config retrieval for the SQLite provider:

1. Check the WAGGLE_SQLITE_DB_PATH environment variable. If not set, default to ~/.waggle/tasks.db.
2. Set the following as the headless_config session variable:
   • dbPath — the resolved path from step 1
   • teamsDatabaseExists (optional — true if teams table has rows)
   • sprintsDatabaseExists (optional — true if sprints table has rows)

Schema Validation

After loading config, verify the database exists and has the correct schema:

sqlite3 "<dbPath>" ".tables"

Expected tables: tasks, task_dependencies, teams, sprints, intake_log.

If any table is missing, run the init script to auto-repair:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/init-db.sh "<dbPath>"

init-db.sh also migrates column additions on an already-initialized database (CREATE TABLE IF NOT EXISTS does not alter an existing table). It runs an idempotent, pragma_table_info-guarded ALTER TABLE ... ADD COLUMN for newer columns such as attachments (the Attachments extended field), so re-running it on any existing DB is safe and a no-op once present.

CRUD Operations

Create Task

Precondition (v2.8.1+): Before invoking the INSERT below, verify that the session-resolved current_user.id is not the fallback sentinel "unknown". If it is, halt and surface an error to the caller:

Cannot create task: current_user.id is "unknown". Configure proper identity resolution before retrying — see the Identity Resolution section below. The simplest fix is to ensure $USER is set in the shell environment, or set WAGGLE_USER_ID explicitly.

This enforces the protocol's "no anonymous tasks" rule. The Identity Resolution section (below) is structured so that id resolves to a real value ($WAGGLE_USER_ID → $USER → "unknown") on every supported environment, so this halt should rarely fire in practice — it catches genuinely unconfigured environments (an unset $USER with no override).

sqlite3 "<dbPath>" "INSERT INTO tasks (title, description, acceptance_criteria, status, priority, executor, requires_review, execution_plan, working_directory, assignee, issuer) VALUES ('<title>', '<description>', '<criteria>', '<status>', '<priority>', '<executor>', <0|1>, '<plan>', '<dir>', '<assignee_json>', '${current_user.id}'); SELECT last_insert_rowid();"

The issuer column receives ${current_user.id} directly from the substituted session variable. The caller does NOT pass an explicit Issuer — per the protocol's Issuer Auto-Populate Contract, Issuer is provider-managed.

To get the generated ID with the minimum required fields, use:

sqlite3 "<dbPath>" "INSERT INTO tasks (title, status, issuer) VALUES ('<title>', 'Backlog', '${current_user.id}') RETURNING id;"

IMPORTANT:

• Escape single quotes in values by doubling them: ' -> ''.
• Apply the same escape to ${current_user.id} if the resolved value can contain quotes (it should not — $USER-derived strings and email addresses are quote-safe by construction, but defensive escaping is recommended).

Update Task

sqlite3 "<dbPath>" "UPDATE tasks SET <field> = '<value>', updated_at = strftime('%Y-%m-%dT%H:%M:%SZ', 'now') WHERE id = '<task_id>';"

For multiple fields:

sqlite3 "<dbPath>" "UPDATE tasks SET status = '<status>', agent_output = '<output>', updated_at = strftime('%Y-%m-%dT%H:%M:%SZ', 'now') WHERE id = '<task_id>';"

tags, assignee, and attachments are stored as JSON-array text. For attachments, set a JSON array of file descriptors — this provider does not host files (supportsFileHosting=false), so each url must be an externally-hosted, caller-supplied URL:

sqlite3 "<dbPath>" "UPDATE tasks SET attachments = '[{\"url\":\"https://files.example.com/spec.pdf\",\"name\":\"spec.pdf\",\"mime_type\":\"application/pdf\"}]', updated_at = strftime('%Y-%m-%dT%H:%M:%SZ', 'now') WHERE id = '<task_id>';"

Get Task

sqlite3 -json "<dbPath>" "SELECT t.*, GROUP_CONCAT(td.blocked_by_id) as blocked_by_ids FROM tasks t LEFT JOIN task_dependencies td ON t.id = td.task_id WHERE t.id = '<task_id>' GROUP BY t.id;"

Delete Task

sqlite3 "<dbPath>" "DELETE FROM tasks WHERE id = '<task_id>';"

Dependencies are automatically removed via ON DELETE CASCADE.

Manage Dependencies (Blocked By)

Add dependency:

sqlite3 "<dbPath>" "INSERT OR IGNORE INTO task_dependencies (task_id, blocked_by_id) VALUES ('<task_id>', '<blocker_id>');"

Remove dependency:

sqlite3 "<dbPath>" "DELETE FROM task_dependencies WHERE task_id = '<task_id>' AND blocked_by_id = '<blocker_id>';"

Querying Tasks

Use the query script for filtered queries with JSON output:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh \
  "<dbPath>" '<where_clause>' '<order_clause>'

Filter Recipes

All tasks (no filter):

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>"

Ready tasks:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "t.status = 'Ready'"

Tasks by executor and status (single executor):

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "t.status = 'Ready' AND t.executor = 'cowork'"

Tasks by executor and status (multiple executors — for cli/claude-desktop environments):

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "t.status = 'Ready' AND t.executor IN ('cli','claude-desktop','cowork')"

Tasks assigned to current user:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "t.assignee LIKE '%<user_id>%'"

Tasks owned by user via Assignee OR Issuer fallback (v2.8.1+):

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" \
  "(t.assignee LIKE '%<user_id>%' OR (t.issuer = '<user_id>' AND (t.assignee IS NULL OR t.assignee = '' OR t.assignee = '[]')))"

Note that t.issuer is a single-value TEXT column (not a JSON array), so it uses = for exact match against <user_id>. This is the SQLite equivalent of the Notion filter Issuer.created_by:{contains:<user_id>}.

In Progress tasks (for concurrency check):

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "t.status = 'In Progress' AND t.assignee LIKE '%<user_id>%'"

Sort by Priority then Due Date:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "" \
  "CASE t.priority WHEN 'Urgent' THEN 1 WHEN 'High' THEN 2 WHEN 'Medium' THEN 3 WHEN 'Low' THEN 4 END ASC, t.due_date ASC"

Hierarchy Queries

Subtasks of a parent:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "t.parent_task_id = '<parent_task_id>'"

Check if a task has children:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "t.parent_task_id = '<task_id>'" | jq '.results | length'

Check if a candidate parent is itself a subtask:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" "t.id = '<candidate_parent_id>'" | jq '.results[0].parent_task_id'

If the result is non-null, the candidate is already a subtask and cannot be used as a parent (2-level limit).

Post-Processing (all queries)

• Blocked By resolved: Check that the blocked_by array is empty OR query each blocked_by task and confirm all have status = 'Done'.
• Sort (if not done in query): Priority — Urgent > High > Medium > Low; then by Due Date (earliest first).

Displaying Task Lists

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" '<where>' '<order>' | \
  jq '[.results[] | {id, title, status, priority, executor, assignee, due_date, blocked_by: (.blocked_by | length | tostring) + " deps"}]'

Task Record Reference

When referring to a task in dispatch prompts and completion instructions, use:

• Task ID: the hex string ID from the id column
• Update instruction: "Run: sqlite3 <dbPath> \"UPDATE tasks SET agent_output = '<result>', status = 'Done', updated_at = strftime('%Y-%m-%dT%H:%M:%SZ', 'now') WHERE id = '<task_id>';\""

On Completion Template

The following template is injected into dispatch prompts by executing-tasks. Placeholders are resolved at dispatch time.

Task ID: <task_id>
Database path: <db_path>

On completion:
1. Run: sqlite3 "<db_path>" "UPDATE tasks SET agent_output='<result>', status='Done', updated_at=strftime('%Y-%m-%dT%H:%M:%SZ','now') WHERE id='<task_id>';"
   - If Requires Review = ON: set status to 'In Review' instead of 'Done'
2. On error: sqlite3 "<db_path>" "UPDATE tasks SET error_message='<error>', status='Blocked', updated_at=strftime('%Y-%m-%dT%H:%M:%SZ','now') WHERE id='<task_id>';"

Pushing Data to View Server

After any task operation (create, update, delete), push fresh data to the local view server:

1. Fetch all tasks:

bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>"

2. Format as TasksResponse and POST:

TASKS_JSON=$(bash ${CLAUDE_PLUGIN_ROOT}/skills/sqlite-provider/scripts/query-tasks.sh "<dbPath>" | jq -c '{tasks: [.results[] | {
  id, title, description, acceptanceCriteria: .acceptance_criteria, status, blockedBy: .blocked_by,
  priority, executor, requiresReview: .requires_review, executionPlan: .execution_plan,
  workingDirectory: .working_directory, sessionReference: .session_reference,
  dispatchedAt: .dispatched_at, agentOutput: .agent_output, errorMessage: .error_message,
  context, artifacts, repository, startDate: .start_date, dueDate: .due_date, tags, parentTaskId: .parent_task_id,
  project, team, assignee, attachments, issuer, url: "", sprintId: .sprint_id, sprintName: null,
  complexityScore: .complexity_score, backlogOrder: .backlog_order
}], updatedAt: (now | strftime("%Y-%m-%dT%H:%M:%SZ"))}')

curl -s http://localhost:3456/api/health -o /dev/null 2>/dev/null && \
  curl -s -X POST http://localhost:3456/api/data \
    -H "Content-Type: application/json" -d "$TASKS_JSON" -o /dev/null 2>/dev/null || true

View Server Field Mapping

SQLite Column	TasksResponse Field
id	id
title	title
description	description
acceptance_criteria	acceptanceCriteria
status	status
blocked_by (via task_dependencies)	blockedBy
priority	priority
executor	executor
requires_review	requiresReview (boolean)
execution_plan	executionPlan
working_directory	workingDirectory
session_reference	sessionReference
dispatched_at	dispatchedAt
agent_output	agentOutput
error_message	errorMessage
context	context
artifacts	artifacts
repository	repository
start_date	startDate
due_date	dueDate
tags	tags (JSON array)
parent_task_id	parentTaskId
project	project
team	team
assignee	assignee (JSON array)
attachments	attachments (JSON array of {url, name, mime_type?, size?}; supportsFileHosting=false — externally-hosted URLs only)
issuer	issuer (single user ID string; auto-populated by Create Task template, v2.8.1+)
(empty string)	url
sprint_id	sprintId
complexity_score	complexityScore
backlog_order	backlogOrder

Identity: Resolve Current User

Called by resolving-identity shared skill when active_provider = sqlite.

SQLite is local — no remote user system. Identity is derived from the shell environment so that multi-user machines and CI environments produce distinct user IDs.

Resolution order:

1. If WAGGLE_USER_ID env var is set and non-empty → use it. This is the override path for environments where $USER is not meaningful (CI runners, shared service accounts, automation).
2. Else if $USER env var is set and non-empty → use it. On Linux / macOS / WSL this gives a per-user shell account name that is unique on the machine. (v2.8.1+: previously this populated only name; now it also populates id.)
3. Else → id ← "unknown". This sentinel signals "identity is genuinely unresolvable" and triggers the Create Task precondition halt.

Concretely set:

• id ← $WAGGLE_USER_ID if non-empty, else $USER if non-empty, else "unknown"
• name ← $USER env var or "unknown"
• email ← null

Note (v2.8.1+): The Create Task precondition halts only when id == "unknown". The literal "local" is no longer used as a fallback — using $USER directly gives a real identifier on every supported environment, eliminating the "every task is owned by 'local'" failure mode.

Identity: Resolve Team Membership

If teams table has rows:

1. Query: sqlite3 -json "<dbPath>" "SELECT * FROM teams;"
2. Parse members JSON array for each team
3. Match by name (case-insensitive) against current_user.name
4. Set current_user.teams and current_team per the same logic as other providers

Identity: List Org Members

SQLite is local — return members from teams table if available, otherwise org_members: [].

sqlite3 -json "<dbPath>" "SELECT members FROM teams;" | jq '[.[].members | fromjson | .[] ] | unique_by(.name)'

Error Handling

Error Category	Condition	Action
Database locked	SQLITE_BUSY	Retryable — wait 1-2 seconds and retry, max 3 attempts
File not found	DB path does not exist	Terminal — instruct user to run setting-up-tasks
Schema mismatch	Missing table or column	Auto-repair — run init-db.sh to create missing tables