Intake

Generate a run id from the current time as MMDD-HHMM. Create the run directory under the user’s current working directory (run pwd and anchor to it) — NOT the plugin install tree. So: <cwd>/.agent-advisor/<run_id>/, plus <cwd>/.agent-advisor/.gitignore containing * (so run state is never committed). All later $RUN_DIR references point here.

Step 1.5 — Temporal signal recognition

If the opening message carries a Temporal signal (mentions Temporal, Temporal Cloud, Temporal workers, temporalio / go.temporal.io / @temporalio / temporal-sdk), acknowledge it in one line and record the signal in $RUN_DIR/context-notes.md. A Temporal migration routes into the normal flow with migrate as the entry point (pre-select it in Q1 when intent is explicit, still confirmable).

No-code path seeding: when the Temporal signal fires AND the user does not provide a code path (no directory offered in Step 4), additionally SEED declared temporal units in $RUN_DIR/context-notes.md: one temporal_worker_poll unit (source: "declared", trigger: "temporal") plus a temporal context stub { "detected": true, "server": "unknown", "source": "declared" }, and note that Activity classes will be established by interview in Clarify.

Continue to Step 2.

Step 2 — Ask two questions with AskUserQuestion

Ask BOTH in one AskUserQuestion call (two questions):

Q1 — Starting point (header “Starting point”):

• Build from scratch — I have an idea, no code → build_scratch
• Deploy existing code — I have working agent code → build_deploy
• Migrate — I have agents running elsewhere → migrate
• Add capabilities — already on AWS, want to add services → add_capabilities

Q2 — Your background (header “Background”):

• Technical (engineer/developer) → technical
• Business-leaning (founder/PM/non-technical) → business
• Mixed team → business (start in business language, add technical detail on request)

Step 3 — Handle add_capabilities

If Q1 == add_capabilities: this is a self-contained branch of THIS skill. Set entry_point = add_capabilities in .phase-status.json (Step 5), then load references/phases/add-capabilities/add-capabilities.md and follow it. Do NOT continue into Discover / Clarify / Design — the branch runs its own 5-step flow and ends by writing capabilities-recommendation.md. (Per the state table, the add_capabilities row routes here directly after intake.)

Step 4 — Open context prompt

Ask (plain text): “What can you tell me about your agent — and is it one workload, or several (e.g. an interactive agent plus a batch job)? If several, do they interact? How is each one triggered (user request, event/queue, cron/schedule)? Any files or existing code to share? (Optional — say ‘skip’ to move on.)” Capture any framework/model/infra hints into $RUN_DIR/context-notes.md. If the user names several workloads, record a draft units list in context-notes.md (one line per unit: proposed kebab-case id, workload_class guess from the closed vocabulary defined in references/phases/discover/discover.md — agent_session | batch | light_io | service — trigger guess from the closed vocabulary (request | event | schedule | temporal | unknown), and interaction). Single-workload answers record nothing extra (collapse invariant — no new question turn either way; this rides the existing question).

Exception — a Temporal signal ALWAYS records a draft, even for a single workload. If the answers mention Temporal (workers, task queues, Activities, Workflows, *.tmprl.cloud), record a draft units list regardless of unit count, and it MUST include a temporal_worker_poll unit plus a line per Activity-execution class the user describes. Do not collapse a Temporal workload to “nothing extra” — the polling tier and the Activity classes are what Clarify’s Temporal interview and scope gate need; losing them silently drops Tier 1 polling and the orchestration decision.

For build_deploy / migrate, explicitly ask for the code path (“Where’s your agent code? A directory path lets me detect your framework/model and skip questions.”). Discover runs only if a path is given; if the user declines a path, note it and set discover = skipped in Step 5.

Step 5 — Write state

Write $RUN_DIR/.phase-status.json with entry_point, audience (from Q2), intake = completed. Set discover = pending if entry point is build_deploy/migrate AND the user offered a code path, else skipped. Set all later phases (clarify, model-recommend, confirm, design, estimate, generate) to pending.

Step 2 amendment — target maturity

Ask Q3 — Target maturity in the same AskUserQuestion call unless a valid run seed supplies target_maturity: Prototype → prototype; Private beta → private_beta; Production → production. Persist the selected value as top-level target_maturity in .phase-status.json. Do not infer this value from a runtime or a successful deployment.