Imported from ianphil/Deming (skills/do/SKILL.md). Install upstream with npx skills add ianphil/Deming --skill do. Copyright stays with the author.

Do

Input

The cycle directory and its ready plan.md.

Process

1. Read the plan and verify the current branch matches its task branch. If the plan is missing, blocked, or ambiguous, return to Plan.
2. Execute only the planned intervention within scope and stop when its acceptance criteria are met. Do not add adjacent cleanup, abstraction, optimization, or documentation; record newly discovered work as a handoff unless the plan is revised through Plan. Collect the specified evidence and observe unexpected effects and operational friction. Check required tools before using or documenting them. Maintain an attempt ledger with a unique evidence ID for every meaningful planned or supplemental attempt: purpose, exact command or artifact, observed result and exit status, what it establishes, and any intervention or limitation. Preserve reproducible checks, their results, and artifact locations in the cycle record. Record failed attempts and their remedies, not just the final passing command; do not delete failed rows after a later retry succeeds.
3. Fill do.md with actual changes, observations, validation commands and results, and deviations. Keep observations separate from interpretations. Label supplemental checks explicitly rather than blending them into planned evidence. Record routine implementation fixes; material changes to the experiment require an explicit stop and handoff rather than silent redesign.
4. Before handing off, verify that the ledger includes every meaningful attempt and that each evidence ID can be referenced by Study.

Output

.deming/cycles/<id>/do.md, marked completed, stopped, or deviated, with evidence and reasons. Reference supporting artifacts when needed; the initial cycle needs no extra evidence folder.

Handoff

Give Study the cycle directory containing both plan.md and do.md. Preserve the original prediction and reserve conclusions for Study.