Tester Write Tests (Fase 2)

Implement coverage from the saved test plan. Write and run tests one by one. Persist the CLI agentSummary markdown for Fase 3.

Prerequisites

• Fase 1 complete: .tester-agent/reports/{runId}/test-plan.md exists.
• Prompt shape: .cursor/skills/tester-agent/templates/prompts/test-implementation.md
• Report contract: .cursor/skills/tester-agent/reports.md
• DoD / examples: .tester-agent/ overrides or bundled templates/workflow/.
• Plan content rules: .cursor/skills/tester-agent/test-design.md

Instructions

1. Read the saved test-plan.md end-to-end.
2. Inspect existing tests now (allowed in this phase only).
3. Derive an ordered work list from the plan: acceptance checks, then each surface's chosen technique (robustness checklist, pairwise rows, valid/invalid transitions, or security cases). Prefer one focused test (or one pairwise row) per iteration.
4. For each item in the work list:
   1. Create or edit only test files (globs: tests/**, **/*.test.*, **/*.spec.*, or project overrides).
   2. Run a focused command for that test only.
   3. On failure, decide at high effort whether the test is wrong or the application missed the acceptance check. Log [fase 2] effort: high.
   4. Test bug → fix the test file → re-run.
   5. App bug vs acceptance → Residual Risk → next (no production edits).
5. Write .tester-agent/reports/{runId}/agent-summary.md with the implementation headings below.
   This file is the skill handoff for CLI field FinalReport.agentSummary (the CLI keeps it in memory; skills need the file).
6. Mark Fase 2 done on the workflow checklist.

Tell the user and append the same line to $REPORT_DIR/progress.log when it happens:

• [fase 2] start: testplan {path}
• [fase 2] gelezen: {path} for the plan and each existing test file opened
• [fase 2] geschreven: {test file}
• [fase 2] commando: {command}
• [fase 2] resultaat: {command} exit {code}
• [fase 2] geschreven: {path to agent-summary.md}

Hard rules

• Only test files this run creates or updates. No production code, config, migrations, Docker, manifests, lockfiles.
• Leave unrelated dirty application files untouched. Do not git restore, git checkout, git reset, git clean, or git stash when a test fails. Fix the test file you wrote, or record residual risk.
• Do not write report.md / report.json / report.csv in this phase.
• Do not update Jira, Google Sheets, or other external systems.
• Implement every planned case: acceptance checks, applicable robustness items, pairwise rows, transitions, and security cases. Skip only rows marked N/A or listed under Not tested.
• Assert the documented error contract (exception vs rejection vs safe default), plus status, response shape, and database side effects when the plan requires them.
• Integration test per acceptance criterion; unit test only for pure helpers.

Required agent-summary.md headings

## Existing Coverage
## Gaps Found
## Test Changes Made
## Commands To Run
## Residual Risk

Under ## Commands To Run, list every concrete shell command Fase 3 should execute (bash code block, one command per line).

One-by-one loop (mandatory)

for each planned test:
  write/update test file(s)
  run focused command
  pass → next
  fail (test bug) → fix test → re-run
  fail (app bug) → residual risk → next

Done when

• Existing coverage compared to plan
• Each planned test written and focused-run (or residual risk recorded)
• agent-summary.md written (CLI-compatible agentSummary body)
• ## Commands To Run ready for Fase 3
• Workflow checklist Fase 2 checked