kanban-worker

Guides Hermes Kanban workers through task lifecycle pitfalls, handoffs, and retry scenarios.

Updated Jul 3, 2026
One-click install
npx skills add https://github.com/CHENHUI-X/toolbox --skill kanban-worker-chenhui-x
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: kanban-worker
Source: https://github.com/CHENHUI-X/toolbox/tree/main/custom-skills/devops/kanban-worker
Command: npx skills add https://github.com/CHENHUI-X/toolbox --skill kanban-worker-chenhui-x

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Dispatched Kanban workers often fail silently or produce unusable handoffs because they mishandle workspaces, write vague block reasons, claim cards they never created, or repeat the mistakes of prior failed runs. This Skill provides the deeper operational detail that prevents those failures. ## Core Features & Use Cases - Workspace and tenant handling: Defines correct behavior for scratch, shared directory, and git worktree workspaces, plus tenant-prefixed memory isolation. - Structured handoffs: Shows good summary and metadata shapes for coding, research, and review tasks, including review-required blocking flows with comment-based context. - Retry diagnostics and pitfalls: Explains how to interpret prior run outcomes (timed_out, crashed, spawn_failed, reclaimed, blocked) and avoid repeating failed approaches. - Use Case: A worker agent spawned to fix a rate limiter completes its work, posts structured metadata as a comment, and blocks with a 'review-required' reason so a human reviewer can approve before merge. ## Quick Start Load this skill when dispatched as a Hermes Kanban worker and follow its guidance to orient on the task, send meaningful heartbeats, and complete or block with well-formed handoffs.

Frequently Asked Questions about kanban-worker

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

FAQPage Schema
How do I complete a Kanban task with structured metadata?

Call kanban_complete with a concise summary and a metadata object containing fields like changed_files, tests_run, and decisions. Shape the metadata so downstream reviewers and aggregators can parse it without re-reading your prose.

How should a Kanban worker handle tasks that need human review?

Post the structured details (changed files, test counts, diff path) as a kanban_comment first, then call kanban_block with a reason prefixed 'review-required:'. The reviewer unblocks the task after approval, which re-spawns you with the comment thread.

What should I do when a Kanban task is a retry of a failed run?

Run kanban_show first and inspect prior runs' outcome, summary, and error fields. Adjust your approach based on the failure type, such as chunking work after a timeout or reducing memory after a crash, instead of repeating the failed path.

Can I claim Kanban cards in created_cards that I did not create?

No. Only pass task ids captured from successful kanban_create return values. The kernel verifies each id, rejects phantom ids with an error, and permanently records the rejected attempt on the task's event log.

Why does the hermes kanban CLI fail inside containerized backends?

The CLI is not installed in Docker, Modal, or SSH terminal backends, so hermes kanban commands fail there. Use the kanban_* tools instead, which work across all terminal backends; the CLI exists for human operators.