Claude Certified Engineers logoCCE

CCAR-P · Study Guide

← Glossary

Domain 7: Developer Productivity & Operational Enablement

Productivity terms: project-scoped config, secret hygiene, workflow gates, outcome metrics, and ops debug.

  • Project-scoped config

    Shared Claude Code rules, skills, and MCP definitions checked into the repo so the team inherits one standard on clone instead of tribal personal setups.

    Exam Personal config standing in for team standards is a red flag.

    See also: 7.1 Configure Claude for teams

  • Secret hygiene

    Keeping API keys and MCP tokens out of VCS, shell history, and prompts — using env vars or a vault, with rotation after any exposure.

    Exam Secrets in git or history fail team-config items.

    See also: 7.1 Configure Claude for teams

  • Bootstrap docs

    A written clone → secrets → verify path that new teammates can follow without Slack lore or copying someone’s home directory.

    Exam Missing bootstrap leaves personal config as the unofficial standard.

    See also: 7.1 Configure Claude for teams

  • AI workflow gates

    Required human or automated checks — especially merge and release — that remain in force while AI drafts reviews, docs, or triage notes.

    Exam Automation that removes necessary review is wrong.

    See also: 7.2 AI-assisted workflows

  • Outcome metrics (dev productivity)

    Time saved and error/revert rates used to judge AI-assisted workflows, as opposed to vanity adoption counts alone.

    Exam Installs without outcome movement do not prove success.

    See also: 7.2 AI-assisted workflows

  • Traces and minimal fixtures

    Request/tool logs plus the smallest reproducible input used to debug intermittent Claude, tool, or MCP failures before rewriting prompts.

    Exam Guessing without traces is the distractor.

    See also: 7.3 Debugging & ops

  • Product vs prompt/tool bug

    Fault isolation that separates wrong business behavior from prompt/skill mistakes, tool/AuthZ failures, and MCP/config wiring issues.

    Exam Collapsing all failures into one bucket fails debug-ops items.

    See also: 7.3 Debugging & ops

  • Ops runbook

    A reusable record of symptoms, signals, repro, classification, fix, and verify steps so on-call does not repeat archaeology for the same agent incident.

    Exam Fix-once-with-no-writeup fails operational enablement.

    See also: 7.3 Debugging & ops