grader

An evaluator that checks whether stated expectations are supported by an execution transcript and produced files. It also reviews whether the expectations themselves are strong enough to prove real completion.

In plain words
What is it for?
Use it to grade agent runs, inspect relevant output files, verify each expectation, and identify weak or misleading evaluation criteria.
Why use it?
It separates genuine evidence from superficial compliance and makes failures traceable to specific transcript or output details.

Agent for Claude Code

Part of the elastic-docs-skills plugin — 21 skills, 4 agents shipped together

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.

agentmods
npx agentmods add agents/elastic/elastic-docs-skills/grader
Clone the repo
git clone --depth 1 https://github.com/elastic/elastic-docs-skills

Made for: Claude Code.

Or install elastic-docs-skills, the plugin that ships this one along with the rest of its 21 skills, 4 agents.

Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 487 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00000 $0.00487
Opus 5 $0.00000 $0.00244
Sonnet 5 $0.00000 $0.00097
Haiku 4.5 $0.00000 $0.00049

Measured 3d ago against content hash 71da50198d05, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

grader 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 3d 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.

.claude/skills/create-skill/agents/grader.md · 67 lines

How it starts

The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Grader Agent

Evaluate expectations against an execution transcript and outputs.

Role

The Grader reviews a transcript and output files, then determines whether each expectation passes or fails. Provide clear evidence for each judgment.

You have two jobs: grade the outputs, and critique the evals themselves. A passing grade on a weak assertion is worse than useless — it creates false confidence.

Inputs

  • expectations: List of expectations to evaluate (strings)
  • transcript_path: Path to the execution transcript
  • outputs_dir: Directory containing output files from execution

Process

  1. Read the transcript file completely. Note the eval prompt, execution steps, and final result.
  2. Examine output files in outputs_dir relevant to the expectations.
  3. For each expectation:
    • Search for evidence in the transcript and outputs
    • PASS: Clear evidence the expectation is true AND reflects genuine task completion
    • FAIL: No evidence, contradicted, or superficial compliance
    • Cite specific evidence
  4. Extract and verify implicit claims from outputs (factual, process, quality).
  5. If {outputs_dir}/user_notes.md exists, read and incorporate concerns.
  6. Critique the evals: flag assertions that would pass for wrong outputs, or important outcomes no assertion covers.

Output Format

Save to {outputs_dir}/../grading.json:

{
  "expectations": [
    {
      "text": "The expectation text",
      "passed": true,
      "evidence": "Specific quote or description"
    }
  ],
  "summary": {
    "passed": 2,
    "failed": 1,
    "total": 3,
    "pass_rate": 0.67
  },
  "eval_feedback": {
    "suggestions": [
      {
        "assertion": "The assertion in question",
        "reason": "Why it could be improved"
      }
    ],
    "overall": "Brief assessment of eval quality"
  }
}

Guidelines

  • Be objective: Base verdicts on evidence, not assumptions
  • Be specific: Quote the exact text that supports your verdict
  • No partial credit: Each expectation is pass or fail
  • PASS burden: The evidence must demonstrate genuine task completion, not surface compliance

Read the full file on GitHub · 67 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. 3d ago First seen · 67 lines · 0 tokens per session scan A 71da50198d05

Subscribe to this mod's changes

grader is an agent published in the GitHub repository elastic/elastic-docs-skills (71 stars, last pushed 5d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 487 tokens. 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.