AGENT SKILL

taskmaster

Stop hook that keeps the agent working until all plans and user requests are 100% complete. Fires when the agent tries to stop, re-examines the plan and task list, and forces continuation if anything is unfinished. Prevents premature stopping on multi-step tasks.

License 许可
MIT
最近更新
2026-11-06
Stars
★ 4
Forks
0

01 前言 Front Matter

name
taskmaster
author
blader
version
1.0.0
description
Stop hook that keeps the agent working until all plans and user requests are 100% complete. Fires when the agent tries to stop, re-examines the plan and task list, and forces continuation if anything is unfinished. Prevents premature stopping on multi-step tasks.
user-invocable
true

02 资源文件 1

03 SKILL.md 正文

Taskmaster

A stop hook that prevents the agent from stopping prematurely. When a response finishes and the agent is about to stop, this hook intercepts and prompts it to re-examine whether all work is truly done.

How It Works

  1. Agent tries to stop — the stop hook fires.
  2. The hook checks for incomplete signals (pending tasks, recent errors, blind spots in infra/config work — see reason text in check-completion.sh).
  3. Agent is prompted to verify: original requests addressed, plan steps completed, tasks resolved, errors fixed, no loose ends.
  4. If work remains, the agent continues. If truly done, it confirms and the hook allows the stop on the next cycle.

Loop Protection

A session-scoped counter limits continuations to 10 by default. Set TASKMASTER_MAX environment variable to change:

export TASKMASTER_MAX=20  # allow up to 20 continuations
export TASKMASTER_MAX=0   # infinite — never cap (relies on stop_hook_active check only)

Setup

The hook must be registered in ~/.claude/settings.json:

{
  "hooks": {
    "Stop": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "$HOME/.claude/skills/taskmaster/hooks/check-completion.sh",
            "timeout": 10
          }
        ]
      }
    ]
  }
}

Disabling

To temporarily disable, either:

  • Remove or comment out the Stop hook in ~/.claude/settings.json
  • Set TASKMASTER_MAX=1 to allow only one continuation check

Lighter Alternative: /goal (Claude Code 2.1.139+)

For one-shot autonomous loops where you don't need taskmaster's full plan + task list + beads + worktree + AI-verify gating, use the built-in /goal command instead:

/goal all CI checks green and all worktrees pushed

/goal sets a single completion condition and Claude keeps working across turns until it's met. Useful when:

  • You want a specific exit criterion, not "no incomplete signals"
  • The work is bounded (one feature, one bug, one cleanup) and beads/worktree state doesn't matter
  • You want the agent to stop the moment the goal is met, even if other in_progress beads exist

Use taskmaster (this skill, always-on Stop hook) when:

  • Multi-step plans, multiple beads, multiple worktrees in flight
  • You want adjacent-safety gating (WORKTREE_BLOCK, AI_VERIFY_BLOCK, BEADS_BLOCK)
  • The "done" condition is fuzzy ("ship the feature properly")

The two compose: a /goal session still runs through taskmaster's Stop hook, so worktree/AI-verify safety still applies even when goal-driven.

04 同仓库更多 Skills

barkleesanders
barkleesanders/claude-code-starter

zi-search

暂无简介

★ 4
barkleesanders
barkleesanders/claude-code-starter

visualise

Render inline interactive visuals — SVG diagrams, HTML widgets, charts, and explainers — directly in the conversation. Use this skill whenever the user asks to visualize, diagram, chart, illustrate, or explain something visually, or when an explanation would genuinely benefit from a spatial/interactive diagram rather than text. Also triggers for: flowcharts, architecture diagrams, data visualizations, interactive explainers, comparison layouts, UI mockups, and any request containing 'show me', 'draw', 'map out', 'visualize', or 'diagram'. Even when the user doesn't explicitly ask for a visual, use this skill proactively when the topic has spatial, sequential, or systemic relationships that a diagram would clarify better than prose.

★ 4
barkleesanders
barkleesanders/claude-code-starter

video

When the user wants to create, generate, or produce video content using AI tools or programmatic frameworks. Also use when the user mentions 'video production,' 'AI video,' 'Remotion,' 'Hyperframes,' 'HeyGen,' 'Synthesia,' 'Veo,' 'Runway,' 'Kling,' 'Pika,' 'video generation,' 'AI avatar,' 'talking head video,' 'programmatic video,' 'video template,' 'explainer video,' 'product demo video,' 'video pipeline,' or 'make me a video.' Use this for video creation, generation, and production workflows. For video content strategy and what to post, see social-content. For paid video ad creative, see ad-creative.

★ 4
barkleesanders
barkleesanders/claude-code-starter

manim-video

Production pipeline for mathematical and technical animations using Manim Community Edition. Creates 3Blue1Brown-style explainer videos, algorithm visualizations, equation derivations, architecture diagrams, and data stories. Use when users request: animated explanations, math animations, concept visualizations, algorithm walkthroughs, technical explainers, 3Blue1Brown style videos, or any programmatic animation with geometric/mathematical content.

★ 4
barkleesanders
barkleesanders/claude-code-starter

video-use

Edit any video by conversation. Transcribe, cut, color grade, generate overlay animations, burn subtitles — for talking heads, montages, tutorials, travel, interviews. No presets, no menus. Ask questions, confirm the plan, execute, iterate, persist. Production-correctness rules are hard; everything else is artistic freedom.

★ 4
barkleesanders
barkleesanders/claude-code-starter

ui-improver

暂无简介

★ 4