nano-spec update

/nano-spec update <task-name> "<what to change>" — reads the existing pack, applies a requested change (mark a todo done, record a decision in doc.md, revise scope, etc.), and appends a dated entry to log.md. This is the mechanism that keeps the pack a living document through the build rather than a write-once spec — the counterpart to nano-spec-create’s one-shot generation.

It maps to no single canonical stage: it is cross-cutting doc-maintenance that runs alongside stage-implement and stage-review as work proceeds. By continuously folding decisions (ADR-style records in doc.mdartifact-adr) and progress (log.md) back into the pack, it applies pattern-session-handoff — the up-to-date pack is what makes the “read all 4 files and explain the project” handoff work.

nano-spec’s log.md journey trail is within-task progress logging, not cross-iteration pattern-knowledge-compounding: it captures learnings for handoff and retrospective, but nano-spec provides no mechanism to harvest them into durable grounding for future tasks (contrast ce-compound / gstack-learn). It therefore stops short of stage-learn.

See Also