# Test Consolidate > Turns overlapping suites into one clean file per source file: BEHAVIOR inventory approved first, originals deleted in the same commit, coverage verified after. TRIGGER WHEN: consolidating, deduping or rewriting the tests of a module, or processing quarantined tests for code being touched. DO NOT TRIGGER WHEN: whole-suite health is measured or quarantined (use /testing:test-audit), or tests written for untested code (use the test-writer agent). - Skill: `acaprino/test-consolidate` (Agent Skill) - Install (CLI): `npx skillmds@latest add acaprino/test-consolidate` - Raw SKILL.md: https://api.skillmd.com/api/skills/acaprino/test-consolidate/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: acaprino (https://skillmd.com/u/acaprino) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/acaprino/test-consolidate --- > Arguments: ` [--runner ] [--coverage-cmd ] [--dry-run]`. Wherever `` appears below, substitute the text the user typed after the skill name. # /testing:test-consolidate Per-module test consolidation: the surgical half of the remediation ladder. Turns N overlapping, contradictory, implementation-coupled test files into one clean file per source file without losing a single behavior that mattered. The behavior inventory comes FIRST and gets approved BEFORE any rewrite; skipping that inventory is how consolidations silently lose the six edge cases the ugly tests existed for. ## Usage ``` /testing:test-consolidate src/auth # consolidate the auth module's tests /testing:test-consolidate src/auth --dry-run # inventory and proposal only /testing:test-consolidate src/billing --coverage-cmd "pnpm vitest run --coverage" ``` ## Arguments - `` (required): the source module whose tests get consolidated. - `--runner ` (optional): run command override when detection would pick wrong. - `--coverage-cmd ` (optional): coverage command override for the baseline and the verification gate. - `--dry-run` (optional): stop after the behavior inventory (Step 2). No writes, no deletions. ## Step 1: Preconditions and baseline 1. Load the `test-hygiene` skill of this plugin; its `references/remediation-workflow.md` section 3 defines this workflow's contract. 2. Verify a git repository with a clean working tree; halt otherwise. 3. Resolve the test set: every test file resolving to `` (imports plus naming convention), wherever it lives, INCLUDING matching entries under `tests/_quarantine/` and their ledger rows. 4. Detect the runner (`--runner` overrides); run the module's tests for a baseline (pass/fail per test). When part of the test set cannot execute locally (missing services, containers, or credentials), say so explicitly, take the baseline for that part from the latest green CI run on the current branch, and record that the Step 7 gate for those tests moves to CI. 5. Record the module's coverage baseline (`--coverage-cmd` or the playbook's per-runner command). No coverage tooling configured: state explicitly that the verification gate degrades from "coverage must not drop" to "count of distinct covered behaviors must not drop", and require the user to acknowledge before proceeding. ## Step 2: Behavior inventory (always, before any code) Read every test in scope and produce the inventory table. Behaviors, not tests: 40 tests commonly reduce to 9 behaviors. | Behavior | file:line | Duplicate of | Value (high/low/none) | Reason | |---|---|---|---|---| Flag separately, each with evidence: - **Contradictory pairs**: tests asserting incompatible outcomes for the same input/state. - **Implementation-coupled**: internal mocks, call-echo asserts, private access. - **Never-failing**: no asserts, tautologies, everything mocked. - **Quarantined entries** for this module, each with a keep (behavior worth preserving in the rewrite) or drop proposal. A drop proposal cites evidence beyond age: feature removed, replacement coverage, temporary origin, no bug-fix provenance. Under `--dry-run`, print the inventory and stop. ## Step 3: Safety-net check Ask the user whether the module sits on a critical business flow. If yes and no e2e test covers that flow (grep the e2e layer), propose writing 1 to 3 e2e tests FIRST and pause consolidation until they pass. Write them per the `developer-essentials:e2e-testing-patterns` skill; it ships in the upstream wshobson/agents marketplace, a hard dependency of this plugin, and if it is unavailable stop and tell the user to install it (`claude plugin marketplace add wshobson/agents`, then `claude plugin install developer-essentials@claude-code-workflows`). The safety net is what makes the deletions in Step 6 safe to approve. ## Step 4: Approval gate Present the inventory via `AskUserQuestion`, grouped per owner (source file for unit tests, behavioral scope above that layer): the keep-list (behaviors the rewrite will cover) and the delete-list (duplicates, never-failing, dropped quarantine entries). Unanswered rows default to KEEP. No flag bypasses this gate. Contradictory pairs need an explicit ruling: which behavior is the correct one (check the production code and its documented contract before proposing). ## Step 5: Rewrite One file per owner at the correct layer, covering exactly the approved behaviors plus any evident gaps the user approved: one test file per source file at the mirrored path for unit tests, one file per behavioral scope for integration, contract, and e2e tests. Follow the prevention rules of the test-hygiene skill; write test content behavior-first per the `mattpocock-skills:tdd` skill (upstream mattpocock/skills, a hard dependency of this plugin; if unavailable, stop and tell the user to install it: `claude plugin marketplace add mattpocock/skills`, then `claude plugin install mattpocock-skills@mattpocock`). ## Step 6: Delete originals, same commit In the SAME commit as the rewrite: delete every original test file in scope, the approved quarantine entries, and their ledger rows. Commit message: `refactor(tests): consolidate , files -> files`. A rewrite commit that leaves the originals alive creates exactly the duplication this command exists to remove. ## Step 7: Verify Run the suite plus the coverage command. Gates: 1. The suite passes. 2. Module coverage is not below the Step 1 baseline (or, in degraded mode, every approved behavior has a covering test). Tests that could not execute locally in Step 1 are gated on CI instead: push, watch the run, and treat a red lane as a failed gate. A gate fails: roll the consolidation commit back, then report which behaviors lost coverage or which tests broke, with the inventory rows involved. The originals come back; nothing is lost. Pick the rollback by push state: `git reset --hard HEAD~1` while the commit exists only locally; `git revert` once it has been pushed, because rewriting history on a shared branch overwrites other sessions' pushes. ## Step 8: Report - Before/after table: files, test cases, runtime, coverage. - Behaviors dropped, each with the user's recorded reason from Step 4. - Quarantine entries processed and entries remaining for other modules. - Suggested next module by `/testing:test-audit`'s latest remediation ranking, when a `TEST_AUDIT.md` exists.