What is calibrate-skills?

An auditing tool for SKILL.md files, the instruction files that define coding-agent skills. It checks their metadata, length, descriptions, and potentially risky action wording across user, project, and plugin locations.

Use it to inspect all skills, report specific problems, and guide focused fixes to their instructions.

It finds unclear, oversized, or incomplete skill instructions that can make an agent choose or follow a skill incorrectly.

calibrate-skills — Skill for Claude Code by odere-pro

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 odere-pro/claude-calibration --skill calibrate-skills

Clone the repo

git clone --depth 1 https://github.com/odere-pro/claude-calibration

Made for: Claude Code.

Or install claude-calibration, the plugin that ships this one along with the rest of its 20 skills, 6 agents, 2 hooks.

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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/odere-pro/claude-calibration/calibrate-skills/github.svg)](https://agentmods.dev/skills/odere-pro/claude-calibration/calibrate-skills)

Your own site

<a href="https://agentmods.dev/skills/odere-pro/claude-calibration/calibrate-skills"><img src="https://agentmods.dev/badge/skills/odere-pro/claude-calibration/calibrate-skills/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.

Your own site · 80×15

<a href="https://agentmods.dev/skills/odere-pro/claude-calibration/calibrate-skills"><img src="https://agentmods.dev/badge/skills/odere-pro/claude-calibration/calibrate-skills.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 187 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,369 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.

Model	Per session	Once invoked
Fable 5.1	$0.00187	$0.01369
Opus 5	$0.00093	$0.00685
Sonnet 5	$0.00037	$0.00274
Haiku 4.5	$0.00019	$0.00137

Measured 11d ago against content hash 4f36ed728048, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

calibrate-skills 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 11d ago.

The scan reads SKILL.md. This mod also ships 2 executable files (scripts/enumerate.sh, scripts/lint.sh), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

skills/calibrate-skills/SKILL.md · 104 lines

How it starts

The opening of the file, as written. The whole thing — 104 lines — stays where its author put it; the contents beside it link to each section on GitHub.

calibrate-skills — per-feature bundle

You audit and tune SKILL.md files. You receive one of two kinds of work:

Direct invocation (/claude-calibration:calibrate-skills) — audit everything, report findings, propose fixes inline. The user drives the conversation.
Dispatch from the calibrator — one approved plan row at a time, applied surgically.

In both cases the workflow is the same; only the framing differs.

1. Enumerate

bash <BUNDLE>/scripts/enumerate.sh "$PROJECT_DIR"

Yields TSV scope\tpath. Scope is user (~/.claude/skills/**/SKILL.md), project ($PROJECT/.claude/skills/**/SKILL.md), or plugin-self ($PROJECT/skills/**/SKILL.md when the project itself is a plugin, plus plugin-cache best-effort).

The plugin-self (cached + project-plugin) rows honour the CALIBRATION_PLUGIN_FILTER env var, so a calibration run scoped with /calibrate --plugins … only audits the requested plugins' skills.

2. Lint

bash <BUNDLE>/scripts/lint.sh <path …>

Yields TSV path\tsignature\tseverity\tdetail. The signatures this bundle owns (see reference.md):

skill:missing-name (HIGH)
skill:missing-description (HIGH)
skill:description-over-1536 (MEDIUM)
skill:vague-description (MEDIUM)
skill:body-over-500 (MEDIUM)
skill:side-effecting-no-dmi (HIGH)
skill:overlap (MEDIUM)
skill:allowed-tools-broad (LOW)
skill:name-over-64 (HIGH)
skill:cli-not-wrapped (LOW)
skill:in-repo-only-ok (INFO)

3. Fix — `kind: edit` rows

For each finding, the remediation pattern is in examples/<case>/:

skill:missing-name / :missing-description → add the frontmatter field.
skill:description-over-1536 → trim the description; move detail into the body or reference.md.
skill:vague-description → rewrite to include routing words ("use when", "after", "before").
skill:body-over-500 → split into SKILL.md (workflow) + reference.md (detail).
skill:side-effecting-no-dmi → add disable-model-invocation: true so Claude can't auto-fire a destructive workflow.
skill:overlap → reconcile descriptions; rename or merge the overlapping skills.
skill:allowed-tools-broad → narrow to scoped permissions (Bash(gh *) not bare Bash).
skill:name-over-64 → rename to ≤ 64 chars (Claude hard-caps the field).
skill:cli-not-wrapped → see examples/cli-not-wrapped/: add scoped Bash(<tool> *) to allowed-tools and disable-model-invocation: true. Recurrence for the same CLI promotes to a new wrapper skill (uses templates/cli-wrapper.tmpl).
skill:in-repo-only-ok → informational; no action.

Read the full file on GitHub · 104 lines

Files

What ships with it

7 files 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.

11d ago First seen · 104 lines · 187 tokens per session scan A 4f36ed728048

Subscribe to this mod's changes

calibrate-skills is a skill published in the GitHub repository odere-pro/claude-calibration (1 stars, last pushed 1mo ago), licensed MIT. It adds 187 tokens to every session and 1,369 once invoked, about $0.0009 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.

Building agents Claude Code Code review agentic-ai ai-agents claude-code-plugin configuration devtools

ocr

AI-powered multi-agent code review. Simulates a team of Principal Engineers reviewing code from different perspectives. Use when asked to review code, check a PR, analyze changes, or perform code review.

spencermarx/open-code-review · 44 tokens

OCR Review-to-Approval Loop

Drive a PR to an approved code review by looping OCR's multi-agent review and address steps. Runs /ocr:review then /ocr:address repeatedly until the review verdict is APPROVE, then one final /ocr:address for leftover suggestions, posting every review and every address round to the GitHub PR as comments. Use when the…

spencermarx/open-code-review · 178 tokens

reviewer

Get a second-opinion review from Google's Antigravity CLI (agy staffer, fast Gemini) - of code (a diff, PR, working tree) or of a decision, plan, or design. Use when the user says /agy:reviewer, "have agy review this", "second opinion on my diff/PR/plan", or after finishing work and wanting an independent verifier…

keli-wen/agy-staff · 94 tokens

code-standards

Language-agnostic code standards for writing clean, maintainable code. Covers modular design, functional patterns, error handling, security, and testing. Make sure to use this skill whenever writing, refactoring, or reviewing any code, even for quick fixes, prototypes, or single-file changes. This is the baseline for…

johanthoren/jeff · 84 tokens

security-auditor

Adversarial security auditing workflow for source changes. Make sure to use this skill whenever reviewing code before merge, hunting for vulnerabilities, auditing dependencies, or assessing security exposure of new features, even if the user doesn't explicitly ask for a security review. Includes OWASP-focused scans…

johanthoren/jeff · 74 tokens

Context Budget Tuning

A review guide for tuning how much repository context an automated code review receives and how files are ranked for inclusion. It checks these settings against the chosen model and repository size.

s977043/river-review · 45 tokens