harness-skill-prune

harness-skill-prune is a skill for Claude Code, Codex from nnabuuu/harness-engineering-toolkit. It costs 199 tokens per session (2,713 once invoked), scanned A, original, MIT.

An audit tool for installed agent skills that decides whether each skill should be retired, shortened, or kept. It treats skills as instruction packages that guide an agent's work.

In plain words
What is it for?
Use it to review a harness's skills, assign each one a disposition, and archive retired skills with a record of why they were removed.
Why use it?
It helps reduce unnecessary instructions, token use, and skills that trigger too often, while preserving organization-specific rules and checks.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one. Also seen: reads .claude/ paths; mentions CLAUDE.md; mentions subagents.

Good fit Use it to review a harness's skills, assign each one a disposition, and archive retired skills with a record of why they were removed.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/nnabuuu/harness-engineering-toolkit/harness-skill-prune
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add nnabuuu/harness-engineering-toolkit --skill harness-skill-prune
Clone the repo
git clone --depth 1 https://github.com/nnabuuu/harness-engineering-toolkit

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for harness-skill-prune

README.md
[![agentmods](https://agentmods.dev/badge/skills/nnabuuu/harness-engineering-toolkit/harness-skill-prune/github.svg)](https://agentmods.dev/skills/nnabuuu/harness-engineering-toolkit/harness-skill-prune)
Your own site
<a href="https://agentmods.dev/skills/nnabuuu/harness-engineering-toolkit/harness-skill-prune"><img src="https://agentmods.dev/badge/skills/nnabuuu/harness-engineering-toolkit/harness-skill-prune/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for harness-skill-prune

Your own site · 80×15
<a href="https://agentmods.dev/skills/nnabuuu/harness-engineering-toolkit/harness-skill-prune"><img src="https://agentmods.dev/badge/skills/nnabuuu/harness-engineering-toolkit/harness-skill-prune.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 199 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,713 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00199 $0.02713
Opus 5 $0.00100 $0.01357
Sonnet 5 $0.00040 $0.00543
Haiku 4.5 $0.00020 $0.00271

Measured 9d ago against content hash 8f2ba4fcfba0, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

harness-skill-prune scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

harness-skill-prune/SKILL.md · 234 lines

How it starts

The opening of the file, as written. The whole thing — 234 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Harness Skill Prune

Skills are engineered capabilities. A skill exists because, at the time it was written, neither the model nor the harness could do the thing natively. Models and harnesses improve; skills don't. Every model generation moves some capabilities from engineered → native, and every skill built on a capability that made that transition turns from an asset into a liability — it still loads, still triggers, still burns tokens, and now competes with a model that does the job better unprompted.

This skill runs the audit and produces a disposition for every installed skill: retire, trim, or keep. Retired skills are archived to a graveyard/ with an epitaph, so the inventory stays honest and the history stays legible.

What survives pruning, structurally: capabilities with no path to native. A model generation can absorb a planning ritual; it cannot absorb your organization's schema, your private domain constraints, or your personal definition of "good." Information asymmetry doesn't expire. Enforcement doesn't expire either — a model can know an invariant and still skip it under pressure, so invariants earn their context by force, not by information. Everything else is on the clock.


Workflow

Run the phases in order. Phases 1–5 are analysis and are always safe: they end in a read-only report, never in file changes. Phase 4 (ablation) is optional but is the strongest evidence you can get. Phase 6 executes the report and runs only after explicit user confirmation — and only after securing a restore path. The default invocation is audit-and-report; "apply" and "restore" are separate, deliberate steps.

Phase 1 — Inventory

Enumerate every installed skill for the target agent(s). Look everywhere skills can load from: project .claude/skills/, user-level ~/.claude/skills/, plugin skill directories, skill registries (skills.json, marketplace manifests), routing tables in CLAUDE.md, and hooks or settings that inject content into every session. For each skill, record:

Read the full file on GitHub · 234 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 9d ago First seen · 234 lines · 199 tokens per session scan A 8f2ba4fcfba0

Subscribe to this mod's changes

harness-skill-prune is a skill published in the GitHub repository nnabuuu/harness-engineering-toolkit (5 stars, last pushed 1mo ago), licensed MIT. It adds 199 tokens to every session and 2,713 once invoked, about $0.0010 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.