Skip to content
Back to skills

Prune No Ops

ASecurity

Delete the sentences in a skill that change no behavior - asides,

  • 4 stars
  • 0 votes
  • 0 copies
  • 4 views
  • Added September 10, 2026
ai-agents

Security analysis

A100/100

Scanned September 10, 2026

npx -y skills add staticaland/skills --skill prune-no-ops --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Prune No Ops?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Prune No Ops
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/staticaland-prune-no-ops/badge)](https://www.skillsdirectory.com/skills/staticaland-prune-no-ops)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: prune-no-ops
description: Delete the sentences in a skill that change no behavior - asides,
  justification, emphasis, and restated defaults.
disable-model-invocation: true
version: 0.1.1
---

# Prune No-ops

Delete every sentence in a target skill that doesn't change behavior. The verdict
for each sentence is the **deletion test**: would an agent running the skill
on its trigger cases behave identically with the sentence and without it? A
sentence whose deletion changes nothing is a **no-op** - delete it whole. A
sentence that survives keeps every word.

Apply the cuts. When the user asks for analysis only, stop after step 2 and
report the verdicts.

## Process

### 1. Read the target

Read the target `SKILL.md` and every file it references. The body prose is the
scope; leave frontmatter untouched - the description's wording does invocation
work, and the deletion test does not grade it. Done when every in-scope file is read.

### 2. Classify every sentence

Run the deletion test on each sentence and record a verdict: _survives_, or one
no-op class:

<!-- vale Skills.WeakVerbs = NO -->
<!-- The bullets quote weak phrasing as the smells they name; "fixing" a
quoted example would erase the smell it demonstrates. -->

- **Aside** - "by the way", "note that", "keep in mind" framing around an
  instruction that binds on its own.
- **Justification** - the why behind an instruction ("this ensures...",
  "because...") when the instruction binds without it. Reasoning survives only
  when it gates a decision: it tells the agent when the rule applies or how to
  trade it off.
- **Default restatement** - an instruction the model already obeys unprompted
  ("be careful", "make sure to read the file first").
- **Emphasis** - "this is important", "always remember", "critically".
- **Duplicate** - a meaning another sentence in the file already states. The
  clearer statement survives. The other is the no-op.

<!-- vale Skills.WeakVerbs = YES -->
<!-- vale Skills.HedgeWords = NO -->
<!-- "A maybe" below is the verdict as a noun, not a hedge. -->

Deletion demands certainty: mark a sentence no-op only when the behavior is
identical either way. A maybe survives. Done when every sentence in scope
carries a verdict.

<!-- vale Skills.HedgeWords = YES -->

### 3. Apply the cuts

Delete each no-op sentence whole. Then reread the pruned file: every surviving
sentence must still bind - repair any reference a cut orphaned ("this",
"it", a dangling "instead"). Done when the file holds only surviving
sentences and each one reads standalone.

### 4. Report

Quote each cut with its no-op class, then the totals: sentences read, cut, and
surviving.

Attribution

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

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.