Tester Run Tests (Fase 3)

Run the relevant tests and write the same three final reports as the CLI.

Prerequisites

• Prefer Fase 2: .tester-agent/reports/{runId}/agent-summary.md
• Else: project testCommands / detected defaults
• test-plan.md path for testPlanPath
• Report contract: .cursor/skills/tester-agent/reports.md
• Definition of Done from .tester-agent/ or bundled template

Instructions

1. Collect commands in order:
   1. Explicit user / project config commands
   2. Else bash block under ## Commands To Run in agent-summary.md
   3. Else detect sensible defaults for the repo
2. Run each command sequentially. Capture command, exitCode, stdout, stderr for every run (needed for report.json / report.md).
3. Determine verdict from test results only:
   • no commands → inconclusive
   • any exitCode !== 0 → failed
   • all 0 → passed
4. Do not check git for changed application code. Do not set policy_violation. The workflow already forbids editing production code; record policy as { "ok": true, "changedFiles": [], "disallowedFiles": [] } so the CLI report shape stays valid.
5. Build the report JSON for the writer. report.json on disk only keeps what report.md shows: runId, createdAt, target, verdict, workflowFiles, summary, policy, and testResults with truncated output. The script drops git, project, workflow contents, testPlanPath, jiraIssueKey, and backlogItemId.
6. Write all three files with the CLI helper (required — do not hand-write formats):

node .cursor/skills/tester-agent/scripts/write-final-report.mjs \
  --dir ".tester-agent/reports/${RUN_ID}" \
  --report-file /tmp/final-report.json

7. Mark Fase 3 and “Final report written” on the workflow checklist.

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

• [fase 3] start
• [fase 3] commando: {command}
• [fase 3] resultaat: {command} exit {code}
• [fase 3] geschreven: {path} for report.md, report.json, and report.csv
• [fase 3] verdict: {passed|failed|inconclusive}

Required outputs

Do not write final-report.md or other alternate names.

Failure notes

When recording failures in agentSummary / residual risk, explain as a tester:

• what did not work
• how to reproduce
• which expected behavior was violated
• which developer action is needed

Hard rules

• Do not edit production code in this phase.
• Do not git restore, git checkout, git reset, git clean, or git stash when a command fails. Record failed and leave the working tree as it is, including the user's other edits.
• Do not mark passed if tests did not run successfully.
• Always use write-final-report.mjs so md/json/csv stay identical to the CLI.

Done when

• All selected commands executed (or blocked with inconclusive)
• Verdict set from test results (passed, failed, or inconclusive)
• report.md, report.json, and report.csv written via script
• Workflow checklist Fase 3 + final report checked