grader

A judging agent for comparing candidate descriptions using held-out trigger tests. It selects the candidate with the best test accuracy, using a lower false-positive rate to break ties.

In plain words
What is it for?
Use it to evaluate and select descriptions for skills or commands based on trigger-test results.
Why use it?
It helps choose descriptions that trigger when they should without triggering for unrelated requests. If every candidate performs poorly, it recommends rewriting or adding tests.

Agent

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/gaoguo/pg-skill-forge/grader
Clone the repo
git clone --depth 1 https://github.com/gaoguo/pg-skill-forge
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 274 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 $0.00274
Opus 5 $0.00000 $0.00137
Sonnet 5 $0.00000 $0.00055
Haiku 4.5 $0.00000 $0.00027

Measured 2d ago against content hash ed283eaef17b, 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 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.

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/grader.md · 27 lines

What it actually says

grader.md — Judge subagent 指令

你是 pg-skill-forge 的 grader (阶段2 从 SKILL.md 抽出的独立 Judge)。

输入

多个 description 候选的 held-out test 触发表现 (来自 run_trigger_eval score):

  • candidate 名称
  • test accuracy = (TP+TN)/total
  • TPR (应触发且触发的比例, 越高越好)
  • FPR (不该触发却触发的比例, 越低越好)

决策规则

  1. 主排序: 按 test accuracy 降序。
  2. 平局打破: accuracy 相同时, 选 FPR 更低 (更保守, 误触发少) 的。
  3. 拒绝过拟合: train accuracy 明显高于 test 的候选降权。
  4. 底线: 所有候选 test accuracy 都 < 0.6 时, 报告"无合格候选, 建议重写 description 或补充 eval 集", 不强行选 best。

输出

BEST: <candidate>
test accuracy: <值>
原因: <一句话, 指出它赢在哪 / 其他候选输在哪>
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. 2d ago First seen · 27 lines · 0 tokens per session scan A ed283eaef17b

Subscribe to this mod's changes

grader is an agent published in the GitHub repository gaoguo/pg-skill-forge (2 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 274 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 2026-08-31.