Working with the agent

Decision delegation

Generate options with their costs, recommend one, keep the approval with the human.

Paid $9 list, $7 until 2026-12-31 · one skill, yours for good. The outline, the steps and the file open after purchase.

Background

When the user hands over the decision, what must be handed back is a traceable recommendation, not a silent unilateral judgement.

Evidence (verbatim): in the same message the user both said "you decide" / "you go ahead and finish everything you can do", and immediately added "after I give my permission, turn it into a new workflow" — the authorization covers execution, not the decision. → profile in references/01-user-decision-profile.md

Why it is built this way

Classification comes before ideation because you decide covers three different scopes. The evidence sits in one user sentence: you decide and finish everything you can do, followed immediately by after I give my permission, turn it into a new workflow. That authorizes execution and leaves the decision with the user. The split therefore runs on reversibility, cost and exposure: delete, overwrite, publish, paid endpoints and anything the user can see wait for one sentence, while naming, directories, script parameters, report wording and execution order get decided and written into the log.

Each candidate carries four columns: approach, cost, irreversibility point, dependency premise. A list without costs hands the comparison back to the user, which is the work the delegation was meant to remove. A recommendation with a fallback and a switch condition turns a pick into a plan. Recording is a script rather than a habit: decision_log.py add takes question, chosen, alternatives, cost, reversible and evidence, and check exits 1 on a missing field, since a record without cost or rationale cannot be compared at the next retrospective.

Where it bites

• Reading you decide as free rein Something irreversible gets decided on the spot: a file deleted, a version published, an outward commitment changed, then revoked and redone. The must-ask row lists delete, overwrite, publish, paid endpoints and anything outward. Gate 1 is enforced in code, where decision_log.py REFUSEs with exit code 2 without the user's own approval.
• Options with no recommendation Two or three candidates get laid out and that is called done, so the user still does the comparison the delegation was supposed to remove. Gate 2 counts that as not a delivery, and step 3 requires which one, one sentence of reason, which one is the fallback, and what condition switches the choice.
• A proposal before the constraints Constraints get skipped and a smooth-looking plan arrives first. The budget question comes back as an afterthought, the plan gets rebuilt, and the round is rework. Gate 3 puts constraint collection ahead of any proposal: money, capability, time, delivery and the hard bounds already given.
• Deciding without a record A choice gets made and the reasoning stays in the conversation. Two rounds later the vendor changes and nobody can say why the first one was picked, so the same decision is made again by feel. Recording is mechanical: decision_log.py add fills the fields and check exits 1 on any missing one.

When to use it

Decision delegation workflow: the four-step process for when the user says 'you decide', 'your call', 'use your judgement', 'give me a recommendation' or 'how should we choose' (collect constraints → generate 2–3 candidates that each carry their cost → rank against the user's preferences and recommend → leave a record + compare afterwards). Core hard rule: what is delegated is generating and recommending; what is retained is approval — irreversible / money-spending / outward-facing decisions must be asked, reversible low-cost ones you decide yourself but must write into the decision log. Use when the user asks 'you decide', 'your call', 'use your judgement', 'go with your recommendation', 'give me a recommendation', 'which one', 'how should we choose' or 'make the decision for me'.