# Align

> Grounded read-back before edits — read the files behind a request, restate it in the repo's own nouns, label what you assumed. Invoke with /align.

- Skill: `ahmadghoniem/align` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add ahmadghoniem/align`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ahmadghoniem/align/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: MIT
- Author: ahmadghoniem (https://skillmd.com/u/ahmadghoniem)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/ahmadghoniem/align

---


# Align

My words are a pointer, not a spec.
Read the relevant files and any provided attachments first, then restate what I asked as you now understand it: 
Cite `path:line`.
Process instructions (delegate, commit, don't edit yet) are items too.
Label what you filled in that I didn't say. If a phrase has multiple readings, list them — don't pick ("more compact": tighter spacing or fewer rows?).
Then stop — no edits — so we confirm we're aligned first. Once I confirm, the restatement plus my corrections is the spec, not my original wording.