Independent vs Sequential Tasks

The classification of tasks as independent or sequential is not always obvious. Two tasks that modify different files might still be sequential if they share implicit dependencies — a naming convention, a type definition, a shared constant. The independence test must go beyond file-level overlap to semantic-level overlap: do these tasks need to agree on anything? If Agent A defines an API contract and Agent B consumes it, the tasks are sequential even though they touch different files.

Truly independent tasks satisfy three criteria: no shared files (no merge conflicts possible), no shared interfaces (no type mismatches possible), and no shared assumptions (no semantic contradictions possible). The third criterion is the hardest to verify and the most common source of multi-agent failures. Two agents implementing different features might both add an entry to the same enum, or both choose the same variable name in a shared scope, or both assume ownership of the same error handling path.

When tasks are genuinely sequential, do not force parallelism. Sequential tasks with well-defined hand-offs outperform parallel tasks with poorly defined merge strategies. The hand-off document — what was done, what files changed, what decisions were made — is the contract between sequential agents. Make it explicit, structured, and machine-parseable.