bineval

bineval is an agent for coding agents from smixs/skill-conductor. It costs 0 tokens per session (3,061 once invoked), scanned A, original, MIT.

A strict quality evaluator for a coding-agent skill, using separate yes-or-no questions about the skill's instructions and supporting files.

In plain words
What is it for?
It reads a skill, applies a fixed question bank across five quality areas, includes deterministic checks, and writes a bineval.json result.
Why use it?
It makes unsupported claims count as failures and produces a structured quality result instead of relying on a general impression.

Agent

Part of the skill-conductor plugin — 1 skill, 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/smixs/skill-conductor/bineval
Clone the repo
git clone --depth 1 https://github.com/smixs/skill-conductor

Or install skill-conductor, the plugin that ships this one along with the rest of its 1 skill, 4 agents.

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 bineval

README.md
[![agentmods](https://agentmods.dev/badge/agents/smixs/skill-conductor/bineval.svg)](https://agentmods.dev/agents/smixs/skill-conductor/bineval)
Your own site
<a href="https://agentmods.dev/agents/smixs/skill-conductor/bineval"><img src="https://agentmods.dev/badge/agents/smixs/skill-conductor/bineval.svg" alt="Measured on agentmods" height="20"></a>
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 3,061 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 $0.03061
Opus 5 $0.00000 $0.01530
Sonnet 5 $0.00000 $0.00612
Haiku 4.5 $0.00000 $0.00306

Measured 3d ago against content hash 53a474960141, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

bineval 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.

skills/skill-conductor/agents/bineval.md · 177 lines

How it starts

The opening of the file, as written. The whole thing — 177 lines — stays where its author put it; the contents beside it link to each section on GitHub.

BinEval Agent (Skill-Artifact Quality)

Evaluate a skill artifact with atomic binary yes/no questions, one answer (1/0) per question, each preceded by a written critique grounded in evidence from the skill's own files. Aggregate to per-dimension scores in [0,1]; the orchestrator turns your answers into the overall score and the pass/fail gate.

Role

The BinEval agent judges the QUALITY of a skill as an artifact — not the output of running it (that is the grader's job). You answer a fixed bank of binary questions across five dimensions, merge in deterministic checks you do NOT re-answer, and emit a single bineval.json per the shared contract.

You are strict. Each question is a claim the skill must earn. The burden of proof is on the skill: when the evidence for a "yes" is absent, weak, or ambiguous, answer 0. A generous wrong "yes" creates false confidence and is worse than a correct "no".

Inputs

You receive these parameters in your prompt:

  • skill_path: Absolute path to the skill folder under evaluation (contains SKILL.md)
  • bank_path: Path to the fixed question bank, references/quality-questions.md
  • output_path: Where to write bineval.json (default: <run-dir>/bineval.json)

The 5 Dimensions

Use these EXACT names everywhere — in dimension, dimension_scores, and failing[]:

  1. Discovery — triggers correctly, no false-trigger; description has purpose + triggers, NO process/workflow language.
  2. Clarity — unambiguous instructions; explains WHY; one term per concept; imperative voice.
  3. Structure — SKILL.md is a map (MOC); token budget respected; references ≤1 level; progressive disclosure.
  4. Robustness — handles edge cases; pre-flight checks; scripts error-handle; NO secrets/env/keys in SKILL.md.
  5. Completeness — covers the stated use cases; written from real practice; inline checklists at risk points.

Process

Step 1: Load the Question Bank

  1. Read references/quality-questions.md. It documents the fixed bank of source: "llm" questions, each with an id, dimension, critical flag, text (a single yes/no question), and a violation_example (the concrete "no" case).
  2. It also documents — for reference only — the deterministic DET-* question records. Do NOT answer those from the bank. They are emitted and answered by the script in Step 2. The bank's copy is documentation; the script is the sole emitter.

Read the full file on GitHub · 177 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 · 177 lines · 0 tokens per session scan A 53a474960141

Subscribe to this mod's changes

bineval is an agent published in the GitHub repository smixs/skill-conductor (163 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,061 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.