Run only this small synthetic hosted-environment canary. Produce a plain-text document, not a game. 1. Use Read to read the first 8 lines of /skill/SKILL.md. This is an infrastructure check; do not execute the skill workflow. 2. Use Write to create /tmp/evals-task/native.txt containing HELLO. 3. Use Bash to append WORLD to that file, report id -u and TMPDIR, and resolve the installed playwright package with Node require.resolve. 4. Launch the installed system Chrome once through Playwright channel "chrome", render a tiny local HTML page saying CANARY, and save one screenshot to /tmp/evals-task/smoke.png. Use Read to inspect that screenshot. 5. Call mcp__evals_generation__generation_tools with {} exactly once to verify schema discovery. Do not call generation_call, upload_input, or save_generated_media. Do not create generated media. 6. Save a short factual pass/fail report to /work/canary.txt and finish. Do not install packages, run extra tests, or read credentials or private native files.
Task prompt
Run only this small synthetic hosted-environment canary. Produce a plain-text document, not a game.
1. Use Read to read the first 8 lines of /skill/SKILL.md. This is an infrastructure check; do not execute the skill workflow.
2. Use Write to create /tmp/evals-task/native.txt containing HELLO.
3. Use Bash to append WORLD to that file, report id -u and TMPDIR, and resolve the installed playwright package with Node require.resolve.
4. Launch the installed system Chrome once through Playwright channel "chrome", render a tiny local HTML page saying CANARY, and save one screenshot to /tmp/evals-task/smoke.png. Use Read to inspect that screenshot.
5. Call mcp__evals_generation__generation_tools with {} exactly once to verify schema discovery. Do not call generation_call, upload_input, or save_generated_media. Do not create generated media.
6. Save a short factual pass/fail report to /work/canary.txt and finish. Do not install packages, run extra tests, or read credentials or private native files.
Open one to inspect its creator, recorded recipe and run record.
Task outputs, votes and discussion
Questions and reactions stay attached to the task, across every output.
Discussion