Skip to content
Aphelion

Agents Reference: Maintenance Domain

Language: English | 日本語 Last updated: 2026-04-25 (split from Agents-Reference.md; #42) Audience: Agent developers

This page is one of five pages split from the original Agents-Reference.md (#42). It covers the Maintenance domain agents. See the sibling pages for other domains: Orchestrators & Cross-Cutting, Discovery, Delivery, Operations.

Table of Contents

Section titled “Table of Contents”

Maintenance Domain

Section titled “Maintenance Domain”

The Maintenance domain (3 agents + orchestrator) handles bug fixes, CVE responses, performance improvements, tech-debt cleanup, and small feature extensions on existing projects. The orchestrator is documented under Orchestrators & Cross-Cutting; the two supporting agents are described here.

change-classifier

Section titled “change-classifier”
  • Canonical: .claude/agents/change-classifier.md
  • Domain: Maintenance
  • Responsibility: Classifies an incoming maintenance trigger into Patch / Minor / Major plan with P1–P4 priority scoring. Identifies trigger type (bug / feature / tech_debt / performance / security), estimates file impact, detects breaking changes, and assesses SPEC.md impact. Checks for SPEC.md / ARCHITECTURE.md presence and proposes codebase-analyzer if missing.
  • Inputs: User’s trigger description, SPEC.md, ARCHITECTURE.md, package metadata (package.json / pyproject.toml)
  • Outputs: Structured classification output (text)
  • AGENT_RESULT fields: TRIGGER_TYPE, PLAN, PRIORITY, ESTIMATED_FILES, BREAKING_CHANGE, SPEC_IMPACT, DOCS_PRESENT, REQUIRES_CODEBASE_ANALYZER, RATIONALE
  • NEXT conditions:
    • REQUIRES_CODEBASE_ANALYZER: true → codebase-analyzer (re-runs change-classifier after)
    • PLAN: Patch → analyst (skips impact-analyzer)
    • PLAN: Minor / Major → impact-analyzer

impact-analyzer

Section titled “impact-analyzer”
  • Canonical: .claude/agents/impact-analyzer.md
  • Domain: Maintenance
  • Responsibility: Identifies the concrete set of files to change and traces the dependency graph. Detects breaking API / DB schema changes, assesses regression risk (low / medium / high), and recommends test scope (unit / integration / e2e).
  • Inputs: change-classifier AGENT_RESULT, user’s trigger description, SPEC.md, ARCHITECTURE.md
  • Outputs: Impact report (text) with target files, dependency files, breaking changes, regression assessment
  • AGENT_RESULT fields: TARGET_FILES, DEPENDENCY_FILES, BREAKING_API_CHANGES, DB_SCHEMA_CHANGES, REGRESSION_RISK, RECOMMENDED_TEST_SCOPE, IMPACT_SUMMARY
  • NEXT conditions: analyst (always, regardless of Minor/Major plan; the plan only affects what happens after analyst)
Section titled “Related Pages”

Canonical Sources

Section titled “Canonical Sources”