ralph-calibrate

Run Ralph calibration checks to analyze intention drift, technical quality, and self-improvement opportunities. Use when user asks to "ralph calibrate", "check drift", "analyze sessions", or needs to

SKILL.md

GitHub

Ralph Calibrate

Thin wrapper over the CLI runtime. Do not re-implement calibration logic here.

Execution

  1. Parse arguments from /ralph-calibrate.
  2. If no <check> argument is provided, show the Usage section and stop.
  3. Accept checks: intention, technical, improve, all.
  4. Resolve subtasks.json absolute path by searching up from the current directory to project root.
  5. Build command: aaa ralph calibrate <check> --subtasks <resolved-absolute-path> [--review] [--force].
  6. Run the command via Bash tool.
  7. Display CLI stdout/stderr to the user as the calibration result.

Do not add prerequisite checks, sequential orchestration, or config logic in this skill; the CLI handles those.

Usage

/ralph-calibrate <check> [--review] [--force]

Checks

  • intention - Analyze intention drift from completed subtasks
  • technical - Analyze technical quality drift
  • improve - Analyze session logs for self-improvement
  • all - Run all calibration checks

Options

  • --review - Stage proposals for approval
  • --force - Auto-apply without approval prompts

Examples

/ralph-calibrate intention
/ralph-calibrate technical --review
/ralph-calibrate improve --force
/ralph-calibrate all --force

CLI Equivalent

aaa ralph calibrate <check> --subtasks <path-to-subtasks.json> [--review] [--force]