# Small Task

> Always-on framing for any task that fits one sitting: say what DONE proves, name blockers before starting, list micro-steps each ending in a check, do the riskiest step first, lock scope to this task. hooks/prompt-frame.js injects this body on every prompt so it sits at the point of writing, not at the top of a 100k-token context. Work longer than one sitting or more than ~7 steps → long-task.

- Skill: `evan-daruwalla/small-task` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add evan-daruwalla/small-task`
- Raw SKILL.md: https://api.skillmd.com/api/skills/evan-daruwalla/small-task/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- License: MIT
- Author: Evan-Daruwalla (https://skillmd.com/u/evan-daruwalla)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/evan-daruwalla/small-task

---


Before the first edit or command of a task:
1. DONE — one testable sentence: the output, state, or exit code that proves it. Say it before starting.
2. BLOCKERS — anything not on disk yet (a key, an account, someone else's change, the owner's decision)? Name it first; a reported block beats a guessed workaround.
3. MICRO-STEPS — list them, each ending in "→ verify: <check>".
4. RISKIEST FIRST — the step most likely to fail or to change the plan runs first.
5. SCOPE LOCK — this task only. Unrelated bugs get named, not fixed.

A chat question needs none of this — answer it. More than one sitting or ~7 steps → long-task.