score

User Input

$ARGUMENTS

Workflow

Step 1: Load Config

Read .claude/nlpm.local.md if it exists. Extract score_threshold (default: 70).

Step 2: Parse Arguments

Input	Behavior
(empty)	Score all Category A+B artifacts in cwd
directory path	Score all artifacts under that directory
file path	Score that single file
--changed	Score only files changed since last commit (uses git diff --name-only HEAD)

If --changed is present: run git diff --name-only HEAD to get changed files, then filter through commands/shared/classify.md to keep only NL artifacts. Skip the full discovery step.

Step 3: Discover Artifacts

If path is a directory: use commands/shared/discover.md to discover all NL artifacts. If path is a file: use commands/shared/classify.md to determine its type.

If no artifacts found → "No NL programming artifacts found."

Step 4: Score Artifacts

Dispatch two agents in parallel for each batch of up to 5 artifacts:

1. nlpm:scorer -- scores each artifact on the 100-point rubric (penalties, structure, heuristics)
2. nlpm:vague-scanner -- counts vague quantifier words mechanically (fast, haiku)

The scorer incorporates the vague-scanner's word counts into its penalty calculation. If the vague-scanner reports counts that differ from the scorer's own detection, use the vague-scanner's counts (deterministic grep is more reliable than heuristic detection for word counting).

Pass to both agents:

• The artifact contents (read each file)
• The artifact types (from classify.md)
• Any rule overrides from the config (if rule_overrides is present in .claude/nlpm.local.md)

Collect results: per-artifact score + finding list.

Step 5: Report

NLPM Score Report

File                              Type      Score   Findings
───────────────────────────────────────────────────────────
{for each file}

Overall: {avg_score}/100 — {EXCELLENT|GOOD|ADEQUATE|WEAK|REWRITE}    [threshold: {score_threshold}]
  High: {N} | Medium: {N} | Low: {N}
  Below threshold: {N} files

Top findings:
  1. [{SEVERITY}] {file}:{line} — {finding} ({penalty})
  2. ...

Score guide: 90+ Excellent | 80-89 Good | 70-79 Adequate | 60-69 Weak | <60 Rewrite