specify-ci

name	specify-ci
description	Specify skill for CI failure investigation — defines analysis categories, skills to load, agents to launch, and severity classification

Spec Filename Prefix

ci-

Skills to Load

• code-follower: Match existing codebase conventions
• code-logic-checker: Find contradictions and logical gaps

Agents to Launch

• fixer: Investigate root causes of independent failures in parallel
• reviewer: Verify root cause analysis is accurate

Analysis Categories

Failure Categories

• Build errors: TypeScript, compilation, or bundling failures
• Test failures: Unit, integration, or e2e test failures
• Lint errors: ESLint, Prettier, or other linting tool failures
• Dependency errors: Install failures, missing packages, version conflicts
• Configuration errors: Workflow YAML issues, missing secrets, environment problems
• Flaky/infra: Timeouts, network errors, runner issues (flag as non-actionable)

Investigation Process

• Fetch CI status: gh run list --branch $(git branch --show-current) --limit 5 --json databaseId,status,conclusion,name,event,headSha
• For each failing run: gh run view <run-id> --json jobs and gh run view <run-id> --log-failed
• Parse logs for error messages, failing steps, exit codes, and stack traces
• For test failures: read the failing test and code under test
• For build/lint errors: trace to source file and identify the violation
• For dependency errors: check package.json, lockfile, and install commands
• For configuration errors: inspect .github/workflows/ YAML files

Severity Classification

• Build-breaking: Prevents any artifact from being produced
• Test failures: Code is broken or tests are wrong
• Lint: Code style violations
• Configuration: Workflow misconfiguration
• Flaky: Non-actionable infrastructure issues

Scope Overrides

• If $ARGUMENTS contains a run ID or workflow name, focus on that specific run
• Otherwise, select all runs with conclusion of failure or startup_failure
• If no failing runs are found, notify the user and stop