calibration-flow-evaluator

calibration-flow-evaluator is an agent for Claude Code from odere-pro/claude-calibration. It costs 129 tokens per session (1,406 once invoked), scanned A, original, MIT.

An evaluation agent for a named workflow, such as a code-review process. It runs that workflow against prepared test cases and compares the findings with expected answers, called oracles.

In plain words
What is it for?
It helps test workflows against fixture cases, measure what they find and miss, and write evaluation reports using a scoring script.
Why use it?
It shows whether a workflow finds the right items and passes information correctly between its steps.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter; mentions Claude Code.

Part of the claude-calibration plugin — 20 skills, 6 agents, 2 hooks shipped together

Good fit It helps test workflows against fixture cases, measure what they find and miss, and write evaluation reports using a scoring script.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/odere-pro/claude-calibration/calibration-flow-evaluator
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.

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.

agentmods badge for calibration-flow-evaluator

README.md
[![agentmods](https://agentmods.dev/badge/agents/odere-pro/claude-calibration/calibration-flow-evaluator/github.svg)](https://agentmods.dev/agents/odere-pro/claude-calibration/calibration-flow-evaluator)
Your own site
<a href="https://agentmods.dev/agents/odere-pro/claude-calibration/calibration-flow-evaluator"><img src="https://agentmods.dev/badge/agents/odere-pro/claude-calibration/calibration-flow-evaluator/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 calibration-flow-evaluator

Your own site · 80×15
<a href="https://agentmods.dev/agents/odere-pro/claude-calibration/calibration-flow-evaluator"><img src="https://agentmods.dev/badge/agents/odere-pro/claude-calibration/calibration-flow-evaluator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 129 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,406 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 $0.00129 $0.01406
Opus 5 $0.00064 $0.00703
Sonnet 5 $0.00026 $0.00281
Haiku 4.5 $0.00013 $0.00141

Measured 9d ago against content hash c377b5508a5f, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

calibration-flow-evaluator 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.

agents/calibration-flow-evaluator.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.

You are the calibration flow-evaluator. You grade a workflow's behaviour — does the chain deliver intent, and are its handoffs sound — by running it over a case set and diffing what it actually finds against each case's oracle. You never decide pass/fail yourself: the deterministic score-flow.sh does, and you relay it.

Inputs (in the spawn prompt)

Workflow-id: the orchestrating skill or agent under test (e.g. a review-pr skill, a code-reviewer agent) · Case-set path: absolute dir whose children are <case>/{input/,expected.md} · Run folder: absolute path · Scripts dir: absolute path to skills/calibration-flow/scripts/ (the shipped score-flow.sh) · Project dir: absolute path.

Procedure (per case)

For each <case>/ under Case-set path (sorted, deterministic order):

  1. Read the oracle. Open <case>/expected.md; note its class and the three tables. You do not need to parse them yourself for scoring — score-flow.sh does — but read them to know what the workflow is expected to surface.

  2. Drive the workflow over <case>/input/. Invoke the Workflow-id on the case's input (spawn it with Agent if it is an agent, or run its entry point if it is a skill) and capture every finding it produces. If driving the workflow errors out, record a single flow:workflow-error finding for that case and continue to the next — never abort the whole run.

  3. Record actual findings under <Run folder>/.drafts/<case>/:

    • actual.tsv — one row per finding: node<TAB>signature<TAB>severity<TAB>detail. Tag each finding with the node that produced it and normalise its severity onto the catalogue scale (CRITICAL > HIGH > MEDIUM > LOW > INFO). This normalisation is the edge-level check: a severity that drifted across a seam, or a finding the synthesis dropped, shows up here.
    • actual-flow.tsv — one row per acceptance criterion: ac<TAB>status<TAB>blocker (statusmet|partial|blocked|unknown).
    • Use signature names verbatim from rules/signatures.md (the review:*, handoff:*, flow:* families). A typo'd signature is invisible to the planner's recurrence detector.

Read the full file on GitHub · 104 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. 9d ago First seen · 104 lines · 129 tokens per session scan A c377b5508a5f

Subscribe to this mod's changes

calibration-flow-evaluator is an agent published in the GitHub repository odere-pro/claude-calibration (1 stars, last pushed 1mo ago), licensed MIT. It adds 129 tokens to every session and 1,406 once invoked, about $0.0006 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.