Terms  /  Execution Quality  /  Task Accountability
04 · Execution Quality

Task Accountability

Every task has a verifiable record of who did what, when, and whether it actually completed.

Task Accountability is provenance for execution. In a multi-agent or human-agent team, knowing that a task is "done" isn't enough — you need to know who did it (which agent, which human), when it was completed, what evidence supports the completion, and whether it was verified. Without Task Accountability, debugging failures becomes a forensic exercise, blame is unassignable, and the same mistakes repeat because nobody can trace them to their source.

Example
A task completion record reads: "Task: Add rate limiting to /api/search. Assigned to: agent:claude-code. Started: 2024-03-14 09:22 UTC. Completed: 2024-03-14 09:41 UTC. Evidence: diff in `middleware/rateLimit.ts`, test output (4/4 passing), manual verification by @mike at 10:15 UTC. Status: verified-complete." Compare this to: "Task: Add rate limiting. Status: done." The first has full accountability. The second is an assertion with no backing.

More in Execution Quality