eval-harness

A framework for testing and measuring coding-agent sessions. It treats evaluations as tests for an AI workflow and checks whether defined tasks meet their success criteria.

In plain words
What is it for?
Use it to define capability and regression evaluations, run them during development, build regression suites, and track pass@k reliability results.
Why use it?
It makes agent behaviour measurable and helps detect regressions when prompts, workflows, or models change.

Skill for Claude CodeCodex

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 skills/t-rav/hydraflow/eval-harness
Any agent
npx skills add T-rav/hydraflow --skill eval-harness
Clone the repo
git clone --depth 1 https://github.com/T-rav/hydraflow

Made for: Claude Code, Codex.

Per session 19 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,405 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin 91% copy Near-identical to another mod 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 $0.00019 $0.01405
Opus 5 $0.00010 $0.00702
Sonnet 5 $0.00004 $0.00281
Haiku 4.5 $0.00002 $0.00140

Measured yesterday against content hash 0bc29828b0b2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

eval-harness 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 yesterday.

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.

Origin

This is a copy

91% identical to eval-harness — 5 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.

.codex/skills/eval-harness/SKILL.md · 237 lines

How it starts

The opening of the file, as written. The whole thing — 237 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Eval Harness Skill

A formal evaluation framework for Claude Code sessions, implementing eval-driven development (EDD) principles.

When to Activate

  • Setting up eval-driven development (EDD) for AI-assisted workflows
  • Defining pass/fail criteria for Claude Code task completion
  • Measuring agent reliability with pass@k metrics
  • Creating regression test suites for prompt or agent changes
  • Benchmarking agent performance across model versions

Philosophy

Eval-Driven Development treats evals as the "unit tests of AI development":

  • Define expected behavior BEFORE implementation
  • Run evals continuously during development
  • Track regressions with each change
  • Use pass@k metrics for reliability measurement

Eval Types

Capability Evals

Test if Claude can do something it couldn't before:

[CAPABILITY EVAL: feature-name]
Task: Description of what Claude should accomplish
Success Criteria:
  - [ ] Criterion 1
  - [ ] Criterion 2
  - [ ] Criterion 3
Expected Output: Description of expected result

Regression Evals

Ensure changes don't break existing functionality:

[REGRESSION EVAL: feature-name]
Baseline: SHA or checkpoint name
Tests:
  - existing-test-1: PASS/FAIL
  - existing-test-2: PASS/FAIL
  - existing-test-3: PASS/FAIL
Result: X/Y passed (previously Y/Y)

Grader Types

1. Code-Based Grader

Deterministic checks using code:

# Check if file contains expected pattern
grep -q "export function handleAuth" src/auth.ts && echo "PASS" || echo "FAIL"

# Check if tests pass
npm test -- --testPathPattern="auth" && echo "PASS" || echo "FAIL"

# Check if build succeeds
npm run build && echo "PASS" || echo "FAIL"

2. Model-Based Grader

Use Claude to evaluate open-ended outputs:

[MODEL GRADER PROMPT]
Evaluate the following code change:
1. Does it solve the stated problem?
2. Is it well-structured?
3. Are edge cases handled?
4. Is error handling appropriate?

Score: 1-5 (1=poor, 5=excellent)
Reasoning: [explanation]

Read the full file on GitHub · 237 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. yesterday First seen · 237 lines · 19 tokens per session scan A 0bc29828b0b2

Subscribe to this mod's changes

eval-harness is a skill published in the GitHub repository T-rav/hydraflow (5 stars, last pushed yesterday), licensed Apache-2.0. It adds 19 tokens to every session and 1,405 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. It is 91% identical to eval-harness, differing in 5 lines, and is treated as a copy.

Related

Other skills, from other repositories

test

Run /test to write a test suite for code you just built or changed, after implementing a feature, route, or fix. Targets uncommitted changes automatically, reads test preferences.json for your framework (asks and saves it if absent), and picks the right strategy per file: happy path, edge cases, error states…

jsmastery-pro/skills · 69 tokens

develop

Run /develop to build a feature, UI or backend, from an approved design, a page, component, API, service, or data slice. If something load bearing is undecided and no spec records it, it stops and routes you to /architect; otherwise it reads the spec plus AGENTS.md, builds, and advances the scope.

jsmastery-pro/skills · 71 tokens

sync

Run /sync as the last step after a change is complete, around merge, to keep durable knowledge current. Updates root and nested AGENTS.md, reconciles the scope from repo evidence, and flags specs the change made stale. Surgical edits only: it adds lines, and rewrites single lines it owns. Never a whole section, never…

jsmastery-pro/skills · 73 tokens

document

Run /document pr | changelog | release-note | postmortem (or let it ask) to write the human facing prose about a change. Drafts from the real commits and diff, writing to the right place. Does not write code, tests, or specs.

jsmastery-pro/skills · 65 tokens

debug

Run /debug to find and fix a bug's root cause: a test failing for an unclear reason, /check verify finding a failure, or behavior being wrong. Runs a reproduce, localize, hypothesize, test, fix, verify loop, makes the minimal fix, and hands a regression test to /test. No features, no extra refactors.

jsmastery-pro/skills · 74 tokens

check

Confirm a change before merge. /check verify drives the real app to prove behavior against the spec (every acceptance criterion met, every surface built). /check review runs a senior code review on a fresh model, one that did not write the code. Verify after /develop, review before a PR. Writes to docs/reviews/, never…

jsmastery-pro/skills · 74 tokens