skill drift

Measures how far a skill has drifted from rubric standards. Each check maps to a drift category — the output tells you where to tighten up.

Usage

doraval skill drift <path> [options]

Drift categories

Category	What it looks for	Pass condition
Trigger	Activation phrases in description	Contains “use when”, “trigger”, or “invoke” (case-insensitive)
Structure	Ordered steps or checklists in body	Matches numbered lists (1.) or bullet lists (-, *)
Voice	Imperative language	Contains verbs like Create, Add, Run, Install, Configure, Set, Build, Use, Check, Verify, Ensure
Example	Code blocks	Contains ``` fenced code blocks
Guardrail	Explicit constraints	Contains MUST or MUST NOT (case-sensitive, uppercase)
Clarity	No ambiguous language	Free of maybe, possibly, consider, you might want to, perhaps

Example

  doraval skill drift — Measuring rubric drift

  Path:  ./skills/my-skill/

  · Trigger    Description includes activation phrases
  · Structure  Has step-by-step instructions
  · Voice      Uses imperative voice ("Do X" not "You might X")
  ↗ Example    No code blocks found — add examples if the skill involves code
  ↗ Guardrail  No explicit constraints — add MUST / MUST NOT guardrails
  · Clarity    No ambiguous language found

  2/6 rubric areas have drifted.

· means the category is clean. ↗ means drift was detected.

Options

Flag	Short	Description
--format <type>	-f	table (default) or json
--agent <name>	-a	Force a specific agent adapter
--verbose	-v	Show detailed diagnostics
--ci		Machine-friendly output, non-zero exit on issues

JSON output

doraval skill drift ./my-skill/ --format json

Returns:

{
  "path": "./my-skill/",
  "driftCount": 2,
  "total": 6,
  "drifts": [
    { "drifted": false, "category": "Trigger", "detail": "Description includes activation phrases" },
    { "drifted": true, "category": "Example", "detail": "No code blocks found — add examples if the skill involves code" }
  ]
}