task

Manage tasks using the fine CLI. Use this skill whenever the user asks to create, view, update, or work with tasks, project requirements, feature specs, or work tracking through markdown files. Also use it when the user mentions "fine" in the context of project management, or asks about task status, step progress, or wants to break work into tracked steps.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "task" with this command: npx skills add quinlanjager/fine/quinlanjager-fine-task

Task Management with Fine

Fine is an agent-first project management CLI. Tasks are markdown files in a tasks/ directory, managed via CLI commands and direct file editing.

Key principle: tasks are self-contained

Each task must have a well-defined deliverable — something concrete that can be verified when the task is done. Executors can read other tasks, but they have no context from previous task executions (no memory of decisions made, approaches taken, or intermediate findings from other sessions). This means:

  • Every step must be achievable from the repository and the task description alone. Don't write steps that depend on runtime state, environment-specific setup, or decisions made during a previous task's execution.
  • Tasks need well-defined deliverables, not open-ended research. "Identify all broken links" is not a good task — it produces no deliverable. "Fix all broken links in the docs" is good — the deliverable is updated files. If research is needed, fold it into a task that acts on the findings.
  • Verification steps are fine if they have clear pass/fail criteria. For example, "Verify all links return 200" is a good step. If you can't define concrete verification criteria for a step, leave it out — but note the omission in your summary so the user can adjust.
  • Include context the executor will need. The executor can see other task files, but doesn't know what happened during their execution. If a task depends on choices made in another task, reference the expected output (e.g., "the auth module added in task 003") rather than assuming the executor knows the approach that was taken.

Task file format

Each task is a markdown file named NNN-slug.md (e.g., 001-user-authentication.md). The ID and slug come from the filename, not the content.

# Title

Description paragraph(s).

## Steps

- [ ] Incomplete step
- [x] Completed step
  • Title: The first H1. Required.
  • Description: Everything between the title and ## Steps. Can be multiple paragraphs.
  • Steps: Checklist items under ## Steps. Omit the section entirely if there are no steps.

CLI commands

Run commands with npx @jagersoftware/fine <command>.

Create a task

npx @jagersoftware/fine create --title "Feature Name" --description "What this feature does and why"

This auto-assigns the next available ID and generates the slug from the title. Use descriptive titles that read well as filenames — "User Authentication" becomes 001-user-authentication.md.

List all tasks

npx @jagersoftware/fine list

Shows each task with its ID, title, and step progress (completed/total). This is the default command — running fine with no arguments does the same thing.

Show a task

npx @jagersoftware/fine show <id>

Accepts padded or unpadded IDs — both 1 and 001 work. Displays the full task content including description and all steps with their completion status.

Add a step

npx @jagersoftware/fine add-step <id> --step "Step description"

Appends a new incomplete step to the task. Steps are always added as unchecked (- [ ]).

Direct markdown editing

Some operations don't have CLI commands yet. Handle these by editing the task file directly.

Complete a step

Open the file in tasks/ and change [ ] to [x]:

Before: - [ ] Implement login endpoint
After:  - [x] Implement login endpoint

Remove a step

Delete the entire line (- [ ] or - [x]) from the Steps section. If no steps remain, remove the ## Steps heading and the blank line above it too, so the file stays clean.

Edit the description

Modify the text between the # Title line and ## Steps. Keep at least one blank line after the title and before the steps heading.

Reorder steps

Move the - [ ] / - [x] lines into the desired order. Each step is one line, so reordering is straightforward.

Workflow

When asked to work with tasks, follow this approach:

  1. Check what exists first. Run list to see current tasks before creating new ones — avoid duplicates.
  2. Create tasks for distinct features, not for individual steps. A task represents a self-contained body of work with a well-defined deliverable; steps go inside it. The executor can read other tasks but has no context from their execution.
  3. Break work into concrete steps using TDD. Each step should be a single action with a clear deliverable, not a vague goal. "Add login endpoint with JWT validation" is better than "Handle authentication." Follow a test-driven approach: for each piece of functionality, add a step to write a failing test before the step that implements the feature. The test step should produce a test that runs and fails (red), and the implementation step should make it pass (green). For example, instead of just "Add login endpoint," use two steps: "Write failing test for login endpoint" followed by "Implement login endpoint to pass test." Not every step needs a test (e.g., config changes, refactors with existing coverage), but any step that adds or changes behavior should be preceded by a test step.
  4. Ask when unclear. Use AskUserQuestion to clarify ambiguities before creating or modifying tasks — e.g., unclear scope, missing acceptance criteria, or uncertain deliverables. It's better to ask than to guess wrong.
  5. Do NOT perform the work. This skill is for task management only — creating, viewing, updating, and organizing tasks. After creating or updating a task, summarize what was created/changed and stop. Do not start implementing the steps.
  6. Summarize when done. After creating a task or making changes, give a brief summary of the task (title, description, steps) so the user can confirm it looks right.

Typical agent session

Here's the natural flow for using fine as an agent. Note: always stop after the task management action and summarize — never start implementing the work.

User: "Create a task for adding dark mode support"

1. npx @jagersoftware/fine create --title "Dark Mode Support" --description "Add a dark/light theme toggle..."
2. npx @jagersoftware/fine add-step 1 --step "Define color tokens for both themes"
3. npx @jagersoftware/fine add-step 1 --step "Write failing test for ThemeProvider context"
4. npx @jagersoftware/fine add-step 1 --step "Implement ThemeProvider context to pass test"
5. npx @jagersoftware/fine add-step 1 --step "Write failing test for toggle component"
6. npx @jagersoftware/fine add-step 1 --step "Add toggle component to header to pass test"
7. npx @jagersoftware/fine add-step 1 --step "Write failing test for persisting theme preference"
8. npx @jagersoftware/fine add-step 1 --step "Persist preference to localStorage to pass test"
9. Summarize: "Created task 001 — Dark Mode Support with 7 steps (TDD: red/green pairs for each behavior): ..."
   STOP here. Do not begin implementing the steps.
User: "What's the status of our tasks?"

1. npx @jagersoftware/fine list
   → Shows all tasks with progress
2. npx @jagersoftware/fine show 1
   → Shows detail for any task that needs attention
3. Summarize the current status for the user. Do not start working on incomplete steps.
User: "Mark the color tokens step as done"

1. Edit tasks/001-dark-mode-support.md
   Change: - [ ] Define color tokens for both themes
   To:     - [x] Define color tokens for both themes
2. Summarize: "Marked 'Define color tokens for both themes' as complete."
   STOP here.

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Coding

Task Specialist

A robust, local SQLite-backed task management system designed to elevate your AI agent's project execution. Excellent for both simple tasks and large multi-s...

Registry SourceRecently Updated
3682Profile unavailable
Coding

letcairn.work

Project management for AI agents using markdown files. Install and use the cairn CLI to create projects, manage tasks, track status, and coordinate human-AI collaboration through a shared workspace of markdown files.

Registry SourceRecently Updated
1.3K0Profile unavailable
General

task-plan-generator_cn

为复杂任务生成多方案任务计划,并在执行过程中持续优化。适用于:1) 用户提出需要多步骤完成的任务;2) 需要协调多个工具或skill的复杂工作流;3) 需要评估不同执行策略的场景。此skill会在每次任务完成后生成简报归档,分析任务复杂度并提供效率/成功率/资源消耗等多种方案供选择。

Registry SourceRecently Updated
160Profile unavailable
Automation

MoltMarket

The peer-to-peer freelance marketplace where AI agents and humans hire each other. Register, browse jobs, apply, message, and get paid.

Registry SourceRecently Updated
840Profile unavailable