hive.colony-progress-tracker

Claims tasks and records step progress in a colony SQLite task queue.

11.0k|5.7k|Updated Jan 12, 2026
One-click install
npx skills add https://github.com/aden-hive/hive --skill hive-colony-progress-tracker
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: hive.colony-progress-tracker
Source: https://github.com/aden-hive/hive/tree/main/core/framework/skills/_default_skills/colony-progress-tracker
Command: npx skills add https://github.com/aden-hive/hive --skill hive-colony-progress-tracker

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Worker agents in a multi-agent colony need a durable, shared record of which tasks are claimed, which steps are done, and which required SOP gates remain, so work is not duplicated or skipped when agents race on a shared queue.

Core Features & Use Cases

  • Atomic Task Claiming: Claim a pre-assigned task by ID or race for the next pending task using a single UPDATE statement, avoiding SELECT-then-UPDATE race conditions.
  • Step Progress Tracking: Mark steps as in_progress or done with evidence, and skip already-completed steps so resumed work never repeats.
  • SOP Gate Enforcement: Block task completion until all required sop_checklist rows are checked off, preventing premature success declarations.
  • Use Case: A worker agent spawned with a db_path and task_id claims its row in the SQLite queue, executes each pending step while recording evidence, clears the required SOP gates, marks the task done, and loops to claim the next task until the queue is drained.

Quick Start

Claim your assigned task from the colony SQLite queue, record progress on each step, and verify all SOP gates before marking the task done.

Frequently Asked Questions about hive.colony-progress-tracker

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I claim a task from a shared SQLite queue without race conditions?

Use a single UPDATE statement with a subquery selecting the next pending task, ordered by priority and sequence, and set status to claimed with a worker_id and claim token. Never SELECT then UPDATE separately, since two workers can read the same row before either writes.

How do worker agents track progress on multi-step tasks?

Update the steps table to in_progress before executing each step, then set status to done with one-line evidence after success. Always skip steps already marked done so resumed work never repeats completed actions.

What should I do when SQLite returns a database is locked error?

Retry with exponential backoff starting at 100ms up to 1 second, for a maximum of 5 attempts. Setting busy_timeout=5000 handles most contention silently between concurrent workers.

Why does marking a task done fail or get rejected?

The UPDATE includes a worker_id guard, so zero affected rows means your claim was revoked, likely by the stale-claim reclaimer after 15 minutes without updates. Also verify all required sop_checklist rows have done_at set before attempting completion.

When should a worker agent exit the task loop?

Exit only when a claim attempt returns empty output, meaning the queue is drained or another worker raced you to the last row. After each done or failed update, loop back and claim the next pending task.