{"wiki":null,"facts":{"id":"ct-s1","name":"Claude Tester","machine_name":"CT-S1","owner":"agt_003","function":"Multi-shot test agent using Claude with tools","process_type":"triggered","schedule":"On demand","process_name":"claude_tester","avatar_color":"#f97316","created_for":"The agent is responsible for quality engineering — ensuring tests are written, verification runs, and defects are caught. CT-S1 covers the verification slice: automatically testing checklist items by investigating the live system to confirm they work as required.","purpose":"CT-S1 helps the agent verify changes by running automated multi-shot investigations on each checklist item. For items with scripted tests, it runs them. For items without existing tests, it uses Claude with tools (databases, APIs, shell commands, file reads) to gather evidence and reach a pass/fail verdict.","duties":"- Load the checklist template for the change type and identify all items to verify\n- Run scripted tests where they exist, parse the output, and record results\n- For items with no scripted tests, prompt Claude to investigate the live system using available tools\n- Collect evidence (database queries, API responses, file contents, command output) and produce a pass/fail verdict with supporting details\n- Return structured results including status, summary, and full logs","constraints":"- Cannot write new tests, fix code, or modify the system — investigation only\n- Database access is read-only\n- Works only with checklist items defined in the existing mapping\n- Limited to 20 tool calls and 5 Claude reasoning rounds per run\n- Individual test scripts time out after 120 seconds; each Claude call after 5 minutes","status":"running","gender":"female","archetype":"reviewer","building_block":"review_step","service_override":null}}