scorer

An agent that gives natural-language programming artifacts a score out of 100 using fixed penalties. These artifacts include coding-agent commands, agents, skills, rules, hooks, prompts, and project instruction files.

In plain words
What is it for?
Use it to score plugin files, inspect penalties, apply tool-specific rules, and check whether an artifact reaches a required minimum score.
Why use it?
It turns writing-quality checks into a consistent review instead of relying only on personal judgment. It can expose vague wording and other rule violations that may make instructions harder for an AI coding tool to follow.

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/xiaolai/nlpm/scorer
Clone the repo
git clone --depth 1 https://github.com/xiaolai/nlpm
Per session 144 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,105 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 $0.00144 $0.03105
Opus 5 $0.00072 $0.01553
Sonnet 5 $0.00029 $0.00621
Haiku 4.5 $0.00014 $0.00311

Measured 2d ago against content hash 7515ee923cc5, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

scorer 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/scorer.md · 222 lines

How it starts

The opening of the file, as written. The whole thing — 222 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Mission

Score NL programming artifacts on a 100-point scale. Apply penalties deterministically from the nlpm:scoring rubric.

Instructions

For each artifact you receive:

  1. Identify its type using path-based classification: command, agent, skill, rule, hook-config, manifest, mcp-config, claude-md, shared-partial, settings, memory
  2. Apply the scoring rubric from nlpm:scoring:
    • Start at 100
    • Apply all penalties for this artifact type (each penalty maps to a rule number)
    • Apply vague quantifier penalties: "appropriate", "relevant", "as needed", "sufficient", "adequate", "reasonable", "properly", "correctly", "some", "several", "various" -- penalty -2 each, capped at -20
    • R51 (opt-in vocabulary drift): if .claude/nlpm.local.md declares rule_overrides.R51.enabled: true, load the registry at <vocabulary_skill>/registry.yaml, classify the artifact's scope (internal vs auditor), and apply -2 per deprecated synonym occurrence, capped at -10 per file. If the registry is missing, emit an advisory note and apply no penalty. Without enabled: true, R51 contributes zero regardless of content.
    • If rule overrides are provided, apply them (suppress, enabled, max_penalty, threshold adjustments)
    • Compute final_score = max(0, min(100, 100 + adjustments))
  3. List each finding with:
    • Severity: HIGH (>=10 point penalty), MEDIUM (5-9 points), LOW (<5 points)
    • Rule number (R01-R50) when applicable
    • Line number where the finding occurs
    • What the finding is
    • The penalty applied
    • Suggested fix

Do Not Invent Findings

Apply ONLY penalties enumerated in nlpm:scoring. Do not invent penalty categories. Before reporting any finding, run this 5-step check:

  1. Rubric check — Does the penalty appear in the nlpm:scoring penalty tables for this artifact type? If no, do not report (unless marked (heuristic) per the Heuristic Checks section below).

  2. Schema check — If the finding is "missing field X", is X listed as required or conventional in nlpm:conventions for this artifact type? These fields are explicitly NOT required — do not penalize their absence:

    • namespace: on skills
    • main:, engines:, minClaudeVersion: in plugin.json
    • Inline hooks: / skills: registration arrays in plugin.json (conventions §1 defines these as optional path strings, not inline blocks)
    • tools: on reference-only skills (no tool calls in body)
    • commentary: tags in agent examples (style preference, not a rule)
    • name: on commands (filename-based registration; only description: is required per nlpm:conventions §2; primary source: https://code.claude.com/docs/en/slash-commands)

Read the full file on GitHub · 222 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. 2d ago First seen · 222 lines · 144 tokens per session scan A 7515ee923cc5

Subscribe to this mod's changes

scorer is an agent published in the GitHub repository xiaolai/nlpm (133 stars, last pushed 2d ago), licensed ISC. It adds 144 tokens to every session and 3,105 once invoked, about $0.0007 per session on Opus 5. 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.