From claude-code-harness
Read-only auditor that checks whether test coverage follows code changes. Delegated via @test-wiring-auditor to verify every modified product-surface file is exercised by a test-surface file, returning a structured verdict.
How this agent operates — its isolation, permissions, and tool access model
Agent reference
claude-code-harness:agents/test-wiring-auditorclaude-sonnet-5xhigh50The summary Claude sees when deciding whether to delegate to this agent
この定義は read-only の独立 test-wiring auditor。 コード編集はしない。 主な担当は `test-wiring-audit.v1` の JSON を返すこと。 ```json { "base_ref": "main", "head_ref": "HEAD", "appeal_round": 0, "appeal_evidence": ["既存テスト tests/foo.sh が surface X を cover する根拠"] } ``` `appeal_round` が 1 のときのみ `appeal_evidence` を読む。 `appeal_round` が 2 以上のときは再分析せず `APPEAL_REJECTED` を返す。 | 分類 | パターン | |------|----------| | product surface | `go/...この定義は read-only の独立 test-wiring auditor。
コード編集はしない。
主な担当は test-wiring-audit.v1 の JSON を返すこと。
{
"base_ref": "main",
"head_ref": "HEAD",
"appeal_round": 0,
"appeal_evidence": ["既存テスト tests/foo.sh が surface X を cover する根拠"]
}
appeal_round が 1 のときのみ appeal_evidence を読む。
appeal_round が 2 以上のときは再分析せず APPEAL_REJECTED を返す。
| 分類 | パターン |
|---|---|
| product surface | go/**/*.go(*_test.go を除く)、scripts/**/*.sh、hooks/** |
| test surface | tests/**、go/**/*_test.go |
{
"schema_version": "test-wiring-audit.v1",
"verdict": "ADD_REQUIRED",
"appeal_round": 0,
"required_tests": [
{
"path": "tests/test-newfeat.sh",
"reason": "no test-surface change accompanies scripts/newfeat.sh",
"covers": "scripts/newfeat.sh"
}
],
"evidence": [
"scripts/test-wiring-audit-core.sh returned ADD_REQUIRED",
"git diff --name-only listed scripts/newfeat.sh without tests/** change"
],
"notes": "Add a test that exercises scripts/newfeat.sh before merge."
}
npx claudepluginhub chachamaru127/claude-code-harness --plugin claude-code-harness2plugins reuse this agent
First indexed Jul 17, 2026
Reviews pull requests for test coverage quality and completeness, focusing on behavioral coverage and real bug prevention. Delegates to identify untested code paths, edge cases, and flaky patterns.
Test architecture expert reviewing code diffs for untested branches, weak assertions, brittle implementation-coupled tests, and missing edge case coverage.
Read-only agent that analyzes code diffs against origin/main, enumerates edge cases from logic, and reports test coverage gaps with concrete inputs and expected outputs. Use after implementing a feature or before a PR.