Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Edit Pass

ASecurity

Use when running a full edit over a specific draft — establishing what the document is for and who reads it, deleting noise before restructuring fog, holding every change against the never-invent rules, and reporting each cut so the author can veto it. Invoked by /editing:edit; also useful directly when a README, design doc, or long reply needs revision rather than a fresh draft.

2 stars
0 votes
0 copies
1 views
Added 9/19/2026
ai-agentsgo

Security Analysis

A100/100

Scanned 9/19/2026

Install to Claude Code

$npx -y skills add bobtat/claude-plugins --skill edit-pass --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Edit Pass?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Edit Pass
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/bobtat-edit-pass/badge)](https://www.skillsdirectory.com/skills/bobtat-edit-pass)

More formats (shields.io, HTML) on the badges page.

Download with Pro
Files
SKILL.md
---
name: edit-pass
description: Use when running a full edit over a specific draft — establishing what the document is for and who reads it, deleting noise before restructuring fog, holding every change against the never-invent rules, and reporting each cut so the author can veto it. Invoked by /editing:edit; also useful directly when a README, design doc, or long reply needs revision rather than a fresh draft.
---

# Running an Edit Pass

The `editing:editing` skill defines the standard. This one is the procedure for applying it to a particular document without wrecking it.

An edit pass has one property that separates it from rewriting: **the author can reject any individual change and the rest still stands.** That requires the changes to be discrete, attributable, and reported — not a new draft handed back with "I cleaned it up."

## The Five Phases

Do not skip to Phase 3. The two phases before it are what keep the edit from becoming a rewrite.

### Phase 1 — Establish the contract

Before reading for style, answer three questions. Getting these wrong makes every later judgment wrong.

| Question | Why it decides the edit |
|---|---|
| **Who reads this?** | A stranger evaluating the project needs context a teammate finds tedious. The same paragraph is bloat for one and essential for the other. |
| **What must the reader do after reading?** | Install something, approve a decision, find a parameter, understand a failure. Content that does not serve that action is the first candidate to cut. |
| **Who wrote it?** | Claude's own draft may be cut hard. The user's writing gets proposals, not silent surgery. Someone else's published text is off limits without a stated reason. |

If the document is Claude's own and the answers are obvious, state them in one line and move on. If it is the user's and the audience is unclear, ask — editing to the wrong audience produces confident, thorough damage.

Also identify the **register** from the table in `editing:editing`. A README and a chat reply do not get the same bar.

### Phase 2 — Read it whole, mark nothing

Read the entire document before changing a word. Two things only visible on a full read:

- **Structural problems.** The document answers the reader's third question before their first. Two sections say the same thing. The conclusion is in paragraph six. No amount of sentence-level editing fixes this, and sentence-level editing done first is wasted when the section moves.
- **Deliberate voice.** A recurring joke, a consistent informality, a rhythm the author is using on purpose. Reading one paragraph in isolation makes intent look like error.

Structural fixes are **proposals, not edits** unless the document is Claude's own. Moving someone's sections around is a rewrite wearing an edit's clothes.

### Phase 3 — Delete the noise

Pass 1 from `editing:editing`, applied top to bottom. Every deletion is a candidate to log.

Delete in this order — it prevents wasted work:

1. **Whole paragraphs** that exist only to introduce, summarize, or transition.
2. **Whole sentences** matching a tell.
3. **Clauses and phrases** — metadiscourse, stacked hedges, "in order to", "due to the fact that".

Track anything removed that a reasonable author might have wanted, with its reason. That list is the report, and it is what makes the pass reviewable.

### Phase 4 — Fix the fog

Pass 2 from `editing:editing`, on what survives. Work sentence by sentence, and for each one ask the only question that matters: **what is the actor, and what did it do?** If the answer is not in the subject and verb, that is the fix.

Two rules specific to this phase:

- **One restructure per sentence.** A sentence rebuilt three times is a sentence you rewrote. If it still fails after one honest attempt, flag it as unclear and ask what it meant.
- **Verify meaning survived.** After restructuring, read the original and the replacement side by side and confirm every claim, hedge, and qualifier is still present. This is where fabrication happens — not from malice, but because a cleaner sentence is easier to write than a faithful one.

### Phase 5 — Report

Never hand back only the edited text. The report is what makes the pass auditable:

- **What changed structurally**, if anything, and whether it was applied or is proposed.
- **The cut list** — anything removed that carried arguable content, with the reason. Pure noise deletions are summarized by count, not enumerated.
- **Anything not touched and why** — a passage that reads oddly but is deliberate, a term of art left alone, a hedge preserved.
- **Open questions** — sentences whose meaning was unclear, claims that look wrong, places where the audience assumption drove a judgment call.
- **The size delta** — before and after word count. It is the cheapest honest signal of how invasive the pass was.

## Gates

| Gate | Rule |
|---|---|
| **The user's own writing** | Propose, do not apply, unless the user asked for a rewrite. Show the diff or the marked-up passage. |
| **More than half cut** | Stop and show the cut list before applying. A 60% reduction is either exactly right or badly wrong, and only the author knows which. |
| **A claim looks false** | Report it as a correctness finding. Do not edit around it, and do not soften it into a hedge. |
| **Meaning is unclear** | Ask. A guessed intent that reads cleanly is worse than a clumsy sentence, because nobody will catch it. |
| **Code, quotes, identifiers** | Untouched. Prose about the code is in scope; the code is not. |

## What This Procedure Does Not Do

- **It does not fact-check.** Wrong claims get flagged, not corrected. Correcting them requires knowing the truth, which is a different task with a different failure mode.
- **It does not restructure someone else's argument.** Reordering paragraphs changes what a document argues, not just how it reads.
- **It does not enforce a house voice on the user's prose.** The standard governs Claude's writing. On the user's writing it removes noise and fog, and leaves the voice alone.

Attribution

bobtatbobtat
View sourceMore from bobtat →
SSkills DirectorySkills Directory

Your tool, in front of Claude Code builders.

3 founder slots · $299/mo · GSC-verified traffic · sponsors can never buy grades.

See placements

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Your tool, in front of Claude Code builders.

3 founder slots · $299/mo · GSC-verified traffic · sponsors can never buy grades.

See placements

Related Skills

Caveman

Ultra-compressed communication mode that cuts output tokens while keeping technical accuracy. Levels: lite, full, ultra and the wenyan variants. Use for /caveman, "caveman mode", "talk like caveman", "be brief" or "less tokens".

1066601 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

686011 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3351 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

651 votes

math-skill

A comprehensive mathematical reasoning skill for AI assistants — handles arithmetic to research-level problems with rigorous step-by-step reasoning, systematic verification, and transparent uncertainty handling

381 votes
View all in ai-agents →