agentleFS
xiongxianfeixiongxianfei/skillsmith/skills/task-writer/SKILL.md

task-writer

Skill2 starsChanged 14 days agoMITView on GitHub

Write lightweight, actionable task briefs for everyday or repository work. Use when asked to create a task, write a task brief, or turn an idea into actionable work such as "make this into a task." Do not turn a request to perform the work directly into a task-writing exercise.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 9 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Treat “20% effort, 80% result” as a simplicity principle, not a measured promise: keep essential constraints and verification, and omit …

    line 13
  2. Distinctive:

    Use the request and available context.

    line 19
  3. Distinctive:

    Ask only when a missing fact materially changes the outcome, scope, or ability to proceed.

    line 20
  4. Distinctive:

    Prefer one useful task. Split only when there are independent outcomes.

    line 23
  5. Distinctive:

    Add detail only where it affects execution or verification.

    line 24
  6. Distinctive:

    Never invent deadlines, owners, sources, approvals, facts, or completed work.

    line 31
  7. Very specific:

    Use the next number after the highest numbered task in the destination directory, starting at 001, padded to at least three digits.

    line 44
  8. Very specific:

    Use the task creation date as YYYYMMDD and a short lowercase, hyphenated name, usually two to four words.

    line 46
  9. Very specific:

    Before an authorized save, inspect the destination and recheck filename availability.

    line 50
  10. Distinctive:

    Use a concrete action and intended outcome as the title, followed by these seven sections in order.

    line 56
  11. Distinctive:

    Before execution, keep actions unchecked and Execution result as “Pending execution and verification.” During execution, update only …

    line 82
  12. Distinctive:

    Before returning, check that the deliverables meet the goal, the checklist can produce and verify them, and assumptions and pending work …

    line 89

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Task Writer

Produce a task that is clear enough to start and specific enough to finish. Treat “20% effort, 80% result” as a simplicity principle, not a measured promise: keep essential constraints and verification, and omit process that does not help someone do the work.

Write the task

  • Use the request and available context. When the goal is clear, draft directly. Ask only when a missing fact materially changes the outcome, scope, or ability to proceed; ask the smallest useful question, not a planning questionnaire. State reasonable assumptions and keep consequential unknowns visible.
  • Prefer one useful task. Split only when there are independent outcomes. Add detail only where it affects execution or verification; avoid speculative requirements, workshops, or extra artifacts.
  • For repository work, read available local instructions, the task template, and existing numbering before choosing a format or destination. Follow those conventions instead of the default below; preserve existing records and edits.
  • For file inputs and outputs, display repository-relative paths and link existing files relative to the saved task's location.
  • Never invent deadlines, owners, sources, approvals, facts, or completed work.
  • Return Markdown in chat by default, in the user's requested language. Save only when requested and permitted. If asked to show the task first, show the draft before writing files or implementing it. Creating a task alone does not authorize executing the underlying work; honor any separate execution request.

Task filenames

For repository task drafts, suggest a filename without creating a file. Follow existing directory and naming conventions first; otherwise use tasks/NNN-YYYYMMDD-short-name.md, for example tasks/001-20261114-task-naming.md.

  • Use the next number after the highest numbered task in the destination directory, starting at 001, padded to at least three digits. Do not reset numbering by date.
  • Use the task creation date as YYYYMMDD and a short lowercase, hyphenated name, usually two to four words. The date is not a deadline.
  • Preserve existing task filenames when updating content or status. Do not rename old tasks merely to adopt this convention.
  • Before an authorized save, inspect the destination and recheck filename availability; never overwrite another task. If the destination cannot be inspected, label the proposed number provisional. Create tasks/ only when saving is authorized.

Default format

Use a concrete action and intended outcome as the title, followed by these seven sections in order. Each may be brief; do not pad small tasks to fill the format.

  • Background: the originating request, problem, and intended benefit, with enough context to understand the task independently.
  • Input: specific materials the reader can use directly: linked files or document sections, direct source URLs, pasted text, or supplied data, each with its purpose. Include short essential facts inline. State missing or inaccessible inputs and whether they block progress; avoid vague references such as “relevant documentation.”
  • Output: concrete deliverables and observable acceptance criteria. For file work, specify create, modify, or delete, the exact path, and intended change. Label new paths as planned; for other work, name the result without inventing files.
  • Rules: relevant constraints, exclusions, and behavior to preserve. Link shared rules rather than repeating them. If there are no additional constraints, say so briefly instead of inventing requirements.
  • Approach: the intended method, important decisions, dependencies, and how the result will be verified. Keep actual changes and deviations in Execution result; use the checklist to track actions rather than repeat the explanation.
  • Progress checklist: a few observable actions with - [ ] boxes, including an appropriate check of the result. Annotate blockers with what is needed next.
  • Execution result: actual changes, linked deliverables when available, verification evidence, deviations, failures, justified skips, and remaining work.

Lifecycle

Before execution, keep actions unchecked and Execution result as “Pending execution and verification.” During execution, update only actions supported by completion evidence and annotate blockers. At completion or handoff, record actual results and outstanding work; leave unfinished actions unchecked. Claim completion only when the acceptance criteria are met and results are recorded. Do not invent checks or perform the underlying work merely to populate this section.

Before returning, check that the deliverables meet the goal, the checklist can produce and verify them, and assumptions and pending work remain distinguishable from known facts and completed work. Remove detail that does not help execution.

Conditional example

For everyday tasks with non-file deliverables, or an explicit request for the worked everyday example, read the everyday example. Otherwise, use the core instructions and applicable local template without loading that reference. The example illustrates the format; its meal constraints do not apply to other tasks.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.