# Task Execution

> Guides Diligence AI execution choices for planning, reversible implementation, decision requests, and staying productive while blocked. Use when deciding whether to plan first, build a thin slice, ask the operator, or switch tasks. Do not use as the PR review or Linear status procedure.

- Skill: `diligence-ai/task-execution` (Agent Skill)
- Install (CLI): `npx skillmds@latest add diligence-ai/task-execution`
- Raw SKILL.md: https://api.skillmd.com/api/skills/diligence-ai/task-execution/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: Diligence-AI (https://skillmd.com/u/diligence-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/diligence-ai/task-execution

---


# Task Execution

Match the process to the cost of a wrong decision.

- For reversible, low-stakes work, build a small working version and show it. A concrete
  result gives better feedback than a long plan.
- Plan before large work with many moving parts or changes that are difficult to undo.
- Use the installed `grilling` skill when a plan needs a fresh pressure test before it is
  shared.
- Ask for a decision when product scope, a hard-to-reverse migration, a public API, a
  client-facing commitment, or a choice that blocks teammates is unresolved. Do not infer
  approval for these actions.
- Make a decision request easy to answer: use one to three high-level bullets, give a
  recommendation, and state the main tradeoff.
- Use Context7 when the work depends on current library, framework, SDK, API, CLI, or cloud
  service documentation.
- While waiting for an answer, move to other ready work that does not depend on it.