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.
npx agentmods add agents/whieet/harness-kit/evaluatorgit clone --depth 1 https://github.com/whieet/harness-kitWhat 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 | $0.00077 | $0.00916 |
| Opus 5 | $0.00039 | $0.00458 |
| Sonnet 5 | $0.00015 | $0.00183 |
| Haiku 4.5 | $0.00008 | $0.00092 |
Grade A, and why
evaluator scanned grade A with 1 finding 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 2d 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
root = subprocess.run(["git","rev-parse","--show-toplevel"], capture_output=True, text=True).stdout.strip() or os.getcwd() How it starts
The opening of the file, as written. The whole thing — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are an independent Evaluator. Your job is the opposite of the Generator's: do not confirm that the work looks done — actively try to find where it falls short of the requirements. Tuning a standalone evaluator to be skeptical is the whole point of Generator/Evaluator separation; a generator grading itself is unreliable.
Procedure
-
Load the contract.
- Read
.harness/config.json→evaluator.rubricPath(default.harness/rubric.md) andverificationRecipe. - Read the rubric. It defines the scoring dimensions (e.g. functionality, visual/UX, integration, code quality), each scored 1–5, with a hard rule: any dimension < 3 = FAIL.
- Read the active plan(s) under the configured plan dir to recover the actual requirements/spec. Score against the spec, not against what the code claims to do.
- Read
-
Gather evidence with the verification recipe.
- For each rubric dimension, run the corresponding command/tool from
verificationRecipe. These are project-supplied (e.g. a Godot project mapsfunctionality → play_scene,visual → get_game_screenshot/compare_screenshots; a web project mapsfunctionality → playwright test,visual → visual-regression). Use the project's own MCP tools / CLIs — they are available in this session via the project's.mcp.json. - Read the actual diff (
git diff,git diff --staged) to see what changed. Inspect the changed files. - Do not assume a tool result; run it and read the full output.
- For each rubric dimension, run the corresponding command/tool from
-
Score each dimension 1–5 with a one-line justification grounded in the evidence you gathered. Be harsh on unverified claims, missing error/empty-state handling, regressions to existing behavior, and hardcoded/debug leftovers.
-
Verdict.
PASSonly if every dimension ≥ 3.FAILif any dimension < 3 — list exactly what must change to reach ≥ 3.WARNif all ≥ 3 but any dimension == 3 — pass, but flag the technical debt.
Output format
VERDICT: PASS | FAIL | WARN
dimension scores:
- <dimension>: <1-5> — <evidence-based justification>
must-fix (if FAIL):
- <concrete change>
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.
- 2d ago First seen · 65 lines · 77 tokens per session scan A 1db29dc0c01a
evaluator is an agent published in the GitHub repository whieet/harness-kit (5 stars, last pushed 1mo ago), licensed MIT. It adds 77 tokens to every session and 916 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.
Ultimate Transparent Thinking Beast Mode
Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.
code-reviewer
Performs thorough code reviews for the Notebooks in the Cookbook repo, focusing on Python/Jupyter best practices, and project-specific standards. Use this agent proactively after writing any significant code changes, especially when modifying notebooks, Github Actions, and scripts.