Problem
An engineering-domain /planning:interview always writes a ## Brief into docs/topics/<topic>/PLAN.md on the task branch. For small ad hoc interviews, such as a scope check inside a larger workflow (one question round before a skill run), that commits a planning file per use. contract_tier: local avoids the commit but still produces the full Brief and ledger.
Surfaced while designing a repo-sweep playbook, where every step asks a short scope round before running a hygiene skill; a Brief per step would add a planning file per step to each PR.
Proposal
A lightweight mode (argument or auto-detected) that runs the rounds and writes no Brief when the caller only needs answers:
- The open-question ledger is still written to the memory slice, and the register gate still applies when questions were asked.
- The decisions are returned to the caller, which records them in its own durable artifact (for example, a
Scope decisions: section in the commit the answers shaped).
Acceptance
- A caller can invoke
/planning:interview for a short scope round and get decisions back with no PLAN.md written or committed.
- The ledger and register gate behave as today.
Problem
An engineering-domain
/planning:interviewalways writes a## Briefintodocs/topics/<topic>/PLAN.mdon the task branch. For small ad hoc interviews, such as a scope check inside a larger workflow (one question round before a skill run), that commits a planning file per use.contract_tier: localavoids the commit but still produces the full Brief and ledger.Surfaced while designing a repo-sweep playbook, where every step asks a short scope round before running a hygiene skill; a Brief per step would add a planning file per step to each PR.
Proposal
A lightweight mode (argument or auto-detected) that runs the rounds and writes no Brief when the caller only needs answers:
Scope decisions:section in the commit the answers shaped).Acceptance
/planning:interviewfor a short scope round and get decisions back with noPLAN.mdwritten or committed.