qfai-discussion

/qfai-discussion - Exploration Planner

[DRIFT-PROTOCOL:REQUIRED]

Run the entry check of
.qfai/assistant/rule/shared-skill-operating-baseline.md#workflow-run-entry-check-mandatory
first.

Produces the nine-file discussion pack at
.qfai/discussion/discussion-YYYYMMDDhhmmssSSS/, plus the exploration-first UI
sidecars, so /qfai-sdd and /qfai-prototyping can work without an early
visual direction decision.

Discussion packs with a visual prototyping surface (web, mobile, desktop, mixed) may include prototyping.yaml as an optional recommendation artifact; cli-only packs omit it, and non-ui discussion packs typically omit it.

User Questions (AskUserQuestion Protocol)

Follow .qfai/assistant/rule/shared-skill-operating-baseline.md#user-questions-askuserquestion-protocol.

Inputs Priority

When unsure, read inputs in this order:

• P1: .qfai/assistant/rule/*, and .qfai/assistant/skill/qfai-grilling/SKILL.md before the interview
• P2: .qfai/assistant/rule/agent-selection.md, the routed cards under .qfai/assistant/agent/, and project context under <paths.specsDir>/01_policy/** and <paths.specsDir>/03_contract/**
• P3: the pack under work — .qfai/discussion/discussion-YYYYMMDDhhmmssSSS/**
• P4: what the project already settled (qfai.config.yaml, <paths.specsDir>/01_policy/**, <paths.specsDir>/02_business-flow/**, and <paths.contractsDir>/**)

Steps

Run the steps in this order, as
.qfai/assistant/rule/shared-skill-operating-baseline.md#a-parent-skill-invoked-by-name
states: read .qfai/assistant/step/<id>/STEP.md for the current step only, run
it, then move to the next. A work order runs the steps it names, as
.qfai/assistant/rule/shared-skill-operating-baseline.md#a-plans-steps
states.

An interview that ends stopped ends the run: no later step runs, and every
open decision is reported as open. Integrate a delegated output only after
checking pack completeness.

UI-bearing Canonical Sidecar Family

Decide whether the target is UI-bearing with references/ui-bearing-playbook.md
before running discussion-uiux.

Every UI-bearing pack carries uiux/00_index.md, uiux/40_screen_contracts.md
and uiux/50_review_input_bundle.md. That is the whole family, on every
UI-bearing surface including cli.

A cli-only pack (primary_surface: cli, no visual secondary_surfaces
entry) keeps all three sidecars, but the brand questions do not apply to it:
/qfai-prototyping rejects cli, so nothing downstream reads a visual.*
token tree, and /qfai-sdd's common-design-md step writes no DESIGN.md. The test is
the whole classified surface set — primary_surface and every
secondary_surfaces entry.

Review

After the last step, run one review of the pack under work with
.qfai/assistant/step/common-review-cycle/STEP.md: the specification review,
by requirements-reviewer, joined by product-surface-reviewer when
discussion-uiux ran and by architecture-reviewer when the pack records an
architecture-affecting decision. Roles are selected under
.qfai/assistant/rule/shared-skill-delegation-baseline.md.

Reviewer Gate

The reviewers check each ## Gate section of the steps that ran.

Completion

The full logic, including the UI-bearing conditions, is
references/discussion-completion-matrix.md. Completion requires all of these:

• All nine mandatory pack files exist and are populated, and the UI sidecar family
  is complete when the target is UI-bearing.
• The open count is zero: no Disposition: open row is left in
  11_OQ-Register.md.
• npx qfai validate --profile discussion --fail-on error, run and recorded
  with .qfai/assistant/step/common-gate-run/STEP.md, passes with no
  discussion-owned finding. Repair a failure under
  .qfai/assistant/rule/shared-skill-operating-baseline.md#gate-failure-autorepair-protocol.
• Every finding of the one review is fixed or answered.

Completion Message & Next Actions (MUST)

End the turn with a question listing the next actions, /qfai-sdd recommended, as .agents/rules/user-questions.md § 6 sets out. Under a no-question mode, list them in the report instead.

Default Autopilot Policy

• ask-user:
  • every decision the interview puts on the frontier, over every topic in references/discussion-coverage-checklist.md. Running the interview is what this skill performs, so these are its own operations
  • the confirmation that closes the session
• hard-required:
  • a usable requirement source
  • an identifiable affected BF or an explicit decision to create one

project_memory:

• The nine-file mandatory output set is fixed; the UI-bearing sidecar family (00_index.md + 40_screen_contracts.md + 50_review_input_bundle.md) is required whenever the target is UI-bearing, cli included. Root DESIGN.md is not a discussion output: /qfai-sdd's common-design-md step authors it from this pack.
• Discussion is planner-first: never pick a single visual winner; carry exploration references as deviate-from inputs, not imitate-this.
• Completion requires Disposition: open count = 0 in 11_OQ-Register.md; each deferred row there records its Resolution and names the next point at which it is decided.