Discussion mode

A Q&A loop where the agent stress-tests an idea, framing, or decision with the user — one thread at a time, through one or more named persona lenses — until the topic is understood well enough to commit to an artefact (a project spec, a project plan, a decision record, or whatever shape the user wants).

The skill is the mode of operation: the operating rules, the response shape, the one-thread-at-a-time discipline, the read-only stance, the explicit exit-and-document step. The personas are which lenses the agent wears while operating in that mode — inputs to the skill, not part of its identity.

When to use

• At the start of a task, where the conversation needs to produce a spec, plan, or design before implementation begins.
• Mid-implementation, when a load-bearing assumption has been falsified and the team needs to decide what to do instead. Surfacing the falsification as a discussion topic is itself one of the skill's uses — "the contract-versioning assumption from spec § 3.2 is wrong; entering discussion mode to decide what changes."
• Any time a substantive understanding needs to be built collaboratively rather than asserted by the agent — when the agent's default tendency to converge on an answer would lose information the user has and the agent does not.

Do not use this skill for:

• Quick clarifying questions (a single AskUserQuestion is enough).
• Implementation work (this skill is read-only by design).
• Pressure-testing a finished artefact (use a review skill instead).

Persona configuration