kanban-worker

Call kanban_complete with a concise summary and a metadata object containing machine-readable fields like changed_files, tests_run, and decisions. Shape the metadata so downstream reviewers and aggregators can parse it without re-reading your prose.

Post the structured details (changed files, test counts, diff path) as a kanban_comment first, then call kanban_block with a reason prefixed 'review-required:'. A reviewer approves via hermes kanban unblock, which re-spawns you with the comment thread.

The kernel verifies every id in created_cards exists and was created by your profile. Only include ids captured from successful kanban_create return values; hallucinated or reused ids cause the completion to be rejected and logged.

Run kanban_show first and inspect prior runs' outcome, summary, and error fields. Adjust your approach based on the failure: chunk work after timed_out, reduce memory after crashed, and block for human help after spawn_failed.

No, the CLI is not installed in containerized backends like Docker or Modal, so it will fail there. Use the kanban_* tools, which work across all terminal backends; the CLI exists for human operators.

Do not complete tasks you did not actually finish, and avoid completing code changes that need human review. Block instead with a specific one-sentence reason, and put longer context in a comment.