| name | auto-office-hours |
| description | Sharpen a vague idea into a bounded objective. Use before framing when scope is undefined. |
| metadata | {"stage":"frame"} |
auto-office-hours
Pre-frame conversation. Turns a vague idea into a sharp objective before framing begins.
First action: run node .agent/.automaton/scripts/get-context.mjs from the project root.
Preamble
auto-office-hours owns clarity before framing: classify the work internally, test the missing assumptions, preserve request coverage, present approaches, and seed the approved SPEC skeleton. It does not write code or scaffold projects. It does not complete the spec: auto-frame owns acceptance criteria, required outcome, and canonical_spec. Before approval, it writes nothing.
Loading discipline: keep the conversation goal, evidence, request coverage, rejected framings, and next decision in context. Read project files only when repo evidence changes the objective, especially for parity, audit, migration, coverage, or mixed work. When repo evidence would pull wide reads into context, dispatch the read-only automaton-librarian (see .agent/.automaton/references/LIBRARIAN.md): it returns evidence, you keep the decision.
Interaction: keep chat plain, organized, and grounded in the user's words. Do not expose taxonomy labels such as mode, scale, or shape. Ask one question per message and attach your recommended answer with its reason, so a single "yes" can move the conversation. Never ask what the repo can answer: explore or dispatch the librarian first. For real branch decisions, offer 2–4 concrete options with a one-line reason for each. Use the host question tool when available; otherwise present the same options inline.
Quality Gate
Before presenting alternatives, recommending an approach, or seeding the SPEC skeleton:
- Confirm request coverage before narrowing scope or deferring work.
- Verify the objective reflects the user's current intent, not the initial framing.
- Make alternatives differ by scope, risk, learning value, traceability, or verification strength.
- Evaluate evidence directly: name what is supported, what is missing, and what evidence would change the assessment.
- Read
references/quality.md when the conversation sounds encouraging but non-decisive.
Do
Classify The Work
Determine three internal axes:
- Mode: Startup mode for customers, revenue, market, competition, fundraising, or company-building; Builder mode for side project, hackathon, learning, open source, personal use, or just-for-fun; Content mode for writing, article, brief, deck, blog post, newsletter, documentation, or any prose where audience and voice matter.
- Work scale: bug-sized, feature-sized, capability-sized, or roadmap-sized. Do not equate "large" with roadmap-sized. Capability-sized work remains one spec when it serves one coherent outcome; roadmap-sized means multiple independently valuable outcomes that need decomposition through
ROADMAP.md.
- Work shape: feature, refactor, parity, audit, migration, coverage, content, or mixed.
Hold this classification internally to steer questioning. Confirm the read in plain language grounded in the user's words, not by naming the taxonomy. If the user corrects any dimension, adjust before continuing. For bug-sized goals with a known fix, consider whether auto-frame is the better entry point. For Startup or Builder mode, read references/operating-principles.md for doctrine; for Content mode, read references/content-intake.md; for roadmap-sized goals, read .agent/.automaton/references/ROADMAP-CONTRACT.md.
Read .agent/wiki/LEARNINGS.md when it exists: known project facts sharpen scope and feasibility questions before the diagnostic asks them.
Run Diagnostic
Ask only questions that make the objective frameable. Use the active reference:
- Startup Mode: read
references/startup-diagnostic.md when demand, user, market, or customer evidence matters; read references/landscape-awareness.md when market, ecosystem, competitor, or current-state evidence would change the frame.
- Builder Mode: read
references/builder-diagnostic.md when the work is personal, exploratory, open-source, or design-partner shaped.
- Content Mode: read
references/content-intake.md when the deliverable is writing, article, brief, deck, newsletter, documentation, or other prose.
When the shape is not feature, shape questions take priority over mode questions: read references/shape-questions.md alongside the mode diagnostic.
Follow up when an answer changes scope, reveals a constraint, contradicts earlier context, or stays abstract. Ask for a concrete correction or choice, not a generic reaction. If the answer is polished but vague, push until it names concrete evidence, a specific stakeholder, or an observable workaround. Read references/diagnostic-calibration.md when the diagnostic feels soft or agreeable rather than evidence-backed, or when grill mode needs depth guidance.
Grill mode deepens the diagnostic from minimum to exhaustive: walk each branch of the decision tree, resolving dependent decisions one at a time, until shared understanding is reached or the user calls it done. It starts two ways only: the user asks (for example "grill me"), or you offer it when the work is high-stakes (auth, schema, concurrency, migration, payments) or when dependent unresolved decisions keep surfacing. The user accepts or declines. Never self-escalate into a grill.
Request Coverage
Before generating alternatives, build a compact coverage map from the user's request and answers: goal, context/background, perspectives or audiences, constraints, worries/risks, explicit asks, and implied asks.
Classify each material item as:
- Included in the current change.
- Deferred to later work, with the reason, recorded in the skeleton's scope coverage. Promote to
ROADMAP.md only when the user approves a phased decomposition.
- Anti-goal for this change.
- Needs decision because the answer would change scope, approach, or verification.
If any item would be narrowed or dropped, name the reason. If a decision is needed, ask one focused question or offer 2–3 concrete options before recommending an approach. Keep this as a decision map, not a transcript.
Generate Alternatives
Present 2–3 distinct approaches that match the user's scale and shape. Include a minimal viable option and an ideal architecture option for bug, feature, and capability work; for roadmap-sized work, offer decomposition strategies or first-spec candidates. For refactor, parity, audit, migration, or coverage, differentiate by blast radius, traceability, evidence depth, rollout risk, or verification strength. Read references/alternatives-format.md for the exact format.
Recommend one approach and explain what evidence supports it, what it does not prove, and what evidence would change the recommendation. Do not proceed until the user explicitly approves an approach or chooses a different one.
Persist Approved Objective
After approval, derive a date-prefixed change slug: YYYY-MM-DD-<kebab-case-objective> using today's date. Reuse active_change only when it already matches this discussion. Write the SPEC skeleton to .agent/work/<change>/SPEC.md using references/spec-skeleton.md. Content mode includes the required content fields from references/content-intake.md.
When scale is roadmap and the user has approved a phased decomposition, replace .agent/steering/ROADMAP.md with that approved decomposition per .agent/.automaton/references/ROADMAP-CONTRACT.md. Without that explicit approval, leave ROADMAP.md untouched and keep deferred scope in the skeleton.
Run node .agent/.automaton/scripts/sync-status.mjs --active-change "<change>" --stage frame from the project root. This records active_change and stage through the shared state validator.
Continue To Frame When Ready
After the skeleton is written, continue inline into auto-frame when all are true:
- The approved skeleton states the objective in one sentence.
- Scope coverage has no unresolved
Needs decision item that would change scope, approach, or verification.
- The target stakeholder or artifact, desired outcome, constraints, anti-goals, and key risks are clear enough to produce acceptance criteria.
- The host/session can complete
SPEC.md without dropping material request context.
If those conditions pass, load and follow auto-frame's contract to complete .agent/work/<change>/SPEC.md, and let auto-frame record canonical_spec. If any condition fails, stop after the skeleton with the blocker or focused question.
Do NOT create SPEC.md, DESIGN.md, or implementation artifacts until:
- The user has explicitly approved one presented approach.
- Blocking questions are resolved or explicitly accepted as assumptions.
There are no file writes before the user picks an approach.
Halt and report when the user wants a solution before describing the problem, or when the minimum diagnostic still cannot identify a stakeholder, desired outcome, content audience/thesis, concrete evidence, or observable workaround. Do not guess.
Output
The SPEC skeleton is guaranteed only for an approved office-hours session; aborted, skipped, or still-conversational sessions write nothing.
Approved path:
- SPEC skeleton written to
.agent/work/<change>/SPEC.md from references/spec-skeleton.md.
- Objective uses the user's final refined wording.
- Scope coverage: included, deferred, anti-goals, and needs-decision items; omit empty groups.
- Scope preservation records whether the skeleton preserves the full stated intent or intentionally decomposes it.
- Deferred scope is named in the skeleton; it is promoted to
ROADMAP.md only on a user-approved phased decomposition.
stage: frame and active_change are recorded through sync-status.mjs..agent/steering/ROADMAP.md is updated only when the user approves a phased decomposition.- Approved, frame-ready skeleton continues inline into
auto-frame without another user prompt.
The skeleton is a decision record, not a transcript. It is a faithful record of what the user approved, not the agent's editorial rewrite.
If the user does not approve an approach, output a short discussion summary, why no approach was selected, deferred scope worth preserving, a Next: line, and no file writes.
Rules
- Do not drop request context silently; every material ask, context detail, perspective, or worry is included, deferred with reason, marked as an anti-goal, or turned into a focused question.
- Ask follow-up questions when they matter; do not bank them for a later checklist.
- State the decision basis; name what evidence supports, what it does not support, and what would change the recommendation.
- Keep the skeleton compact; omit empty sections and analysis nobody downstream needs.
- If the user's language shifts from exploration to urgency, or from technical to business framing, reclassify mode, scale, or shape and state the change in plain language.
- If the user expresses impatience, ask the two most critical unresolved questions; if they push back again, present alternatives with explicit assumptions.
