metric-design

metric-design is a skill for Claude Code, Codex from kirodotdev/KiroCrew. It costs 75 tokens per session (652 once invoked), scanned A, original, Apache-2.0.

A first-stage process for designing a trustworthy measurement for improving a software project. It defines the main number to track, supporting stages, reference results, safety limits, and checks against misleading improvements.

In plain words
What is it for?
Use it to inspect a repository, create and calibrate a measurement configuration, record baselines, and set rules that prevent regressions before optimization begins.
Why use it?
An optimization loop can appear to improve a project when its measurement is noisy or measures the wrong thing.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one.

Good fit Use it to inspect a repository, create and calibrate a measurement configuration, record baselines, and set rules that prevent regressions before optimization begins.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/kirodotdev/kirocrew/metric-design
About the project

Kiro Crew is a persistent development workspace where agents continue multi-step software work across sessions, schedules, and connected interfaces. Developers use it locally or remotely through a desktop app, web dashboard, CLI, Slack, or Discord, with unattended tasks and recurring jobs. The catalogue contains skills and instructions for working with this workspace.

kirodotdev/KiroCrew · 3,687 stars · on GitHub · kiro.dev

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 kirodotdev/KiroCrew --skill metric-design
Clone the repo
git clone --depth 1 https://github.com/kirodotdev/KiroCrew

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 metric-design

README.md
[![agentmods](https://agentmods.dev/badge/skills/kirodotdev/kirocrew/metric-design.svg)](https://agentmods.dev/skills/kirodotdev/kirocrew/metric-design)
Your own site
<a href="https://agentmods.dev/skills/kirodotdev/kirocrew/metric-design"><img src="https://agentmods.dev/badge/skills/kirodotdev/kirocrew/metric-design.svg" alt="Measured on agentmods" height="20"></a>
Per session 75 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 652 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 $0.00075 $0.00652
Opus 5 $0.00037 $0.00326
Sonnet 5 $0.00015 $0.00130
Haiku 4.5 $0.00007 $0.00065

Measured 7d ago against content hash 678a0cde32ac, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

metric-design 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 7d 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.

src/kiro_crew/apps/builtins/auto_improvement/skills/metric-design/SKILL.md · 52 lines

How it starts

The opening of the file, as written. The whole thing — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.

metric-design — build the ruler (Phase 1)

Build the ruler before you measure anything with it. This skill drives Phase 1: analyze the target repository, then design a metric the keep-or-revert loop can trust. A loop optimizing a noisy or wrong ruler "wins" on fiction, and that is the dominant risk this app is built to eliminate — so Phase 2 does not start until Phase 1 has proven itself.

The output — the ruler, as durable config plus a reviewable doc

Writes data/ruler/: the calibrated ruler config and a human-readable metric-design document to review before any optimization runs. The ruler the active target profile supplies:

  • Primary metric — a low-variance number where the two measurement arms cancel as much shared cost as possible (label + unit + direction). Never hard-coded in the UI; it is read from the profile.
  • Attributable sub-stages — so a win is pinned to a named stage rather than hand-waved as a whole-system improvement.
  • Frozen anchors — reference measurements with provenance (for example a pinned floor plus a shipped-defaults baseline), recorded in data/ruler/.
  • Guardrails — metrics that must not regress beyond a stated tolerance.
  • Reward-hack guards — checks the build/test gate structurally cannot see, such as "no silent capability shrink" or "a held-out functional probe still passes".

Calibration — the trust gate

  1. Noise band — around 30 repetitions of the untouched baseline under the full harness, then noise_band = max(2σ, floor). Any delta inside the band is no change, not a small win.
  2. Canary (mandatory) — a known or deliberately forced win that MUST clear the band. If it cannot, the harness is broken and the run halts. This is the Phase-1 gate: no Phase-2 cycle is trusted until the canary passes. The UI disables Start until the ruler reports status: "calibrated", and the backend independently refuses to run on an uncalibrated ruler — two checks because a UI-only guard is bypassable.

Read the full file on GitHub · 52 lines

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. 7d ago First seen · 52 lines · 75 tokens per session scan A 678a0cde32ac

Subscribe to this mod's changes

metric-design is a skill published in the GitHub repository kirodotdev/KiroCrew (3,687 stars, last pushed today), licensed Apache-2.0. It adds 75 tokens to every session and 652 once invoked, about $0.0004 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 2026-08-30.