kanban-worker

Executes kanban tasks in isolated workspaces and hands off results via structured completion or blocking.

6|1|Updated May 11, 2026
One-click install
npx skills add https://github.com/yakeworld/Synthos --skill kanban-worker-yakeworld
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: kanban-worker
Source: https://github.com/yakeworld/Synthos/tree/main/skills/extended/external-automation/devops/kanban-worker
Command: npx skills add https://github.com/yakeworld/Synthos --skill kanban-worker-yakeworld

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agents working on kanban tasks often lose context, leak data across tenants, or falsely claim completion, leaving boards inconsistent and downstream workers without usable handoffs. This Skill enforces disciplined task execution and structured handoff inside the Hermes kanban system. ## Core Features & Use Cases - Workspace-aware execution: Handles scratch, shared directory, and git worktree workspace kinds, including automatic git worktree add initialization when .git is missing. - Structured handoffs: Uses kanban_complete with machine-readable metadata, kanban_block with review-required: prefixes for code changes needing human review, and kanban_comment as the durable annotation channel. - Integrity gates: Verifies created_cards IDs against actual kanban_create return values, rejects phantom IDs, and diagnoses retry scenarios from prior run outcomes (timeout, crash, spawn failure). - Use Case: An agent picks up a task to implement a rate limiter, works in a git worktree, runs 14 tests, posts structured metadata as a comment, and blocks the task with a review-required reason so a human can approve before merge. ## Quick Start Pick up my assigned kanban task, check its status with kanban_show, complete the work in the workspace, and hand it off with a structured summary and metadata.

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 as an autonomous agent?

Run kanban_show first to confirm the task is not blocked or archived, do the work in $HERMES_KANBAN_WORKSPACE, then call kanban_complete with a concise summary and structured metadata. For code changes needing review, use kanban_block with a review-required reason instead.

When should I use kanban_block instead of kanban_complete?

Use kanban_block when the task needs a human decision or code review, prefixing the reason with 'review-required: ' for code changes. Put structured metadata like changed files and test counts in a kanban_comment first, since block only carries a human-readable reason.

How does the kanban worker handle git worktree workspaces?

If the workspace kind is worktree and .git does not exist at the resolved path, run git worktree add from the main repository first, then cd in and commit work normally. Scratch workspaces are disposable; dir: workspaces are shared persistent state.

Why was my kanban_complete rejected for created_cards?

The kernel verifies every ID in created_cards exists and was created by your profile. Only pass IDs captured from successful kanban_create return values; phantom or failed-create IDs trigger a gate rejection recorded permanently on the task event log.

How do I prevent memory leaks across tenants in kanban tasks?

When $HERMES_TENANT is set, prefix every persistent memory entry with the tenant namespace, such as 'business-a: Acme is our biggest customer'. Unprefixed entries leak context across tenant boundaries.

What should I do when retrying a failed kanban task?

Inspect the prior runs' outcome and error fields from kanban_show before acting. Timeouts suggest chunking the work, crashes suggest reducing memory, spawn failures indicate profile config issues best escalated via kanban_block rather than blind retries.