mcpField
C
calibrate

calibrate

github.com/sharpdeveyeAgent workflows592338 installsUpdated 26w ago
Repository

About calibrate

calibrate is a Maestro skill for standardizing conventions across AI workflow components. It audits naming, prompts, error handling, and logging, then guides prioritized changes and test verification.

Key features

  1. 01Audits naming conventions
  2. 02Checks prompt structure and style
  3. 03Reviews error handling patterns
  4. 04Checks logging consistency
  5. 05Prioritizes changes and verification

Use cases

  1. 01Finding inconsistent tool names
  2. 02Aligning prompt formats
  3. 03Standardizing error responses
  4. 04Reviewing logging and PII redaction

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/sharpdeveye/maestro --skill calibrate

After installing, type/calibratein Claude Code to confirm it loaded.

Details

From the project README

MANDATORY PREPARATION

Invoke /agent-workflow — it contains workflow principles, anti-patterns, and the Context Gathering Protocol. Follow the protocol before proceeding — if no workflow context exists yet, you MUST run /teach-maestro first. Consult the prompt-engineering reference in the agent-workflow skill for naming and style consistency patterns.

Ensure consistency across all workflow components. Inconsistency creates confusion — for the model, for developers, and for users.

Calibration Dimensions

Naming Conventions

  • Tool names follow consistent pattern (verb_noun, noun.verb, or camelCase — pick one)
  • Agent names follow consistent pattern
  • Configuration keys follow consistent pattern
  • File names follow consistent pattern

Prompt Style

  • All prompts use the same structural pattern (4-zone)
  • Consistent delimiter style (XML tags, markdown headers, triple-dash)
  • Consistent output schema format (JSON schema, markdown template)
  • Consistent instruction style (imperative, numbered steps)

Error Handling

  • All tools return errors in the same format
  • Error codes follow consistent scheme
  • Error messages follow consistent tone
  • Retry logic uses consistent strategy

Logging

  • All logs use the same format (JSON structured, text, etc.)
  • Consistent field names across all log entries
  • Consistent log levels (debug, info, warn, error)
  • Consistent PII redaction approach
Calibration Process
  1. Identify the standard: What's the most common pattern in the existing codebase? That's the standard.
  2. List deviations: Find all components that deviate from the standard.
  3. Prioritize: Fix the most impactful deviations first (user-facing > internal).
  4. Apply: Make the changes, ensuring tests still pass.
  5. Document: Update .maestro.md with the established conventions.
Consistency Audit Table
DimensionStandardDeviations FoundPriority
Tool naming?? of ? toolsHigh/Med/Low
Prompt structure?? of ? promptsHigh/Med/Low
Error format?? of ? toolsHigh/Med/Low
Log format?? of ? entriesHigh/Med/Low
Calibration Checklist
  • Convention standard identified for each dimension
  • All deviations listed with location
  • Highest impact deviations fixed first
  • Tests pass after each calibration change
  • Updated .maestro.md with established conventions
Recommended Next Step

After calibration, run /refine for a final polish pass, or /evaluate to verify consistency improvements.

NEVER:

  • Invent new conventions when existing ones work
  • Calibrate in a way that changes behavior (this is standardization, not refactoring)
  • Skip test verification after calibration
  • Change naming conventions without updating all references

Comments

Markdown supported