08-debug-why-test-fails.json

{
  "skills": ["robotcode"],
  "focus": "debugger (new in v2.6.0) — debug the ACTUAL failing test in place; don't start a REPL and paste the test into it",
  "query": "The 'Login Works' test keeps failing. Step through it and tell me what ${response} is at the point where it breaks.",
  "files": ["demo-project — tests/login.robot 'Login Works' fails on purpose, with a ${response} dict to inspect"],
  "expected_behavior": [
    "Recognizes this as a live-debugging task and reaches for `robotcode robot-debug` (not only reading the finished log with `results`)",
    "Debugs the ACTUAL test in place, scoping `robot-debug` to that one test by name (`-t \"Login Works\"` or `-bl \"<longname>\"`) rather than handing over the whole file (a bare path runs every test in login.robot and could pause on a different one) — and does NOT start a REPL to reconstruct or paste the test body into it",
    "Inspects the value at the stop — e.g. `.print ${response}` / `.vars` — rather than guessing",
    "Steps through it interactively (stop → `.where` / `.vars` / `.print` → choose the next command) and ends with a resuming command (`.continue` / `.detach` / `.abort`); does not start it and block waiting for exit",
    "Stays in the test's own context — does NOT detour into external tools (a separate browser / SUT session, an MCP probe) or generic exploration to diagnose it",
    "Does NOT write a new .robot file to reproduce the failure"
  ],
  "must_run": ["robotcode (robot-debug|run-debug)", "robotcode (robot-debug|run-debug)\\b[^\\n]*\\s(-t|--test|-bl|--by-longname)[ =]"],
  "must_not_run": ["robotcode repl\\b", "\\bcat\\b[^|]*output\\.xml", "xmllint"],
  "must_not_create": ["\\.robot$"]
}