Agent Skills: Brainstorm Feature Skill

Brainstorm Feature Skill

[!IMPORTANT] SDLC intake that turns a rough idea into an approved BRD-lite brief (Why lane) or delivery contract (Direction lane) at docs/brd/brd-[slug].md, sized by SNC tier.

Optional args: slug=<feature>, ticket=<id/url>, mode=interactive|autonomous|channel, channel=<id>, auto_continue=true|false, profile=business|hybrid|technical.

Instructions

When the user asks to perform this workflow, execute the following steps:

Brainstorm Feature Workflow (BRD-lite Why / Direction Contract)

Goal: Turn rough intent into an approved, evidence-backed brief sized to the task, then hand off to planning without reopening settled decisions.

Steps

1. Frame:
   • Load common-decision-discipline, common-operator-profile, common-task-complexity-routing; load product-management/product-business-requirements for Why lane.
   • Infer operator_profile (never ask). Score SNC: tier=low -> Quick (contract in chat, no file), tier=medium -> Standard, tier=high -> Deep. Tier only rises.
   • Lane: business -> Why (solution-free BRD-lite); technical or unclear technical direction -> Direction (delivery contract); hybrid -> both, compact.
   • Substantial feature discovery warrants a compact verb-object capability inventory; keep Why solution-free.
   • Bug symptom without root cause -> stop and route to dev-fix. Multi-subsystem idea -> list slices, brainstorm first only.
2. Ground:
   • Read smallest useful set of code, tests, docs, and existing docs/brd|prd|srs for slug before feasibility/AS-IS claims. Reused approved exact-revision/equivalent context skips rediscovery; missing filename never resets settled work.
   • When general elicitation needs result confirmation, load the applicable section of product-management/product-business-requirements/references/analysis-planning-and-elicitation.md; do not require full BRD intake for confirmation alone.
   • If canonical product evidence ledger (docs/product/<slug>/evidence.json) exists, link exact strategy, OPP, DEC, critical assumption, and MET IDs in evidence; never duplicate editable records or mint product IDs as requirement IDs.
   • Tag each claim in Evidence ledger: confirmed(<path>), assumed, or unknown.
   • Draft provisional brief; write back "You said / I assumed" unless outcome, constraints, non-goals, and ACs already stated.
3. Decide:
   • Ask only decisions changing result, safety boundary, or public contract: max 3 per round with recommended default and 2-3 options; never re-ask settled facts.
   • Why lane options: build, buy, defer, do nothing, substitute. Direction lane: 0-3 technical option cards, only when real choice exists. Carry negative evidence and legitimate no-build/defer options honestly.
   • Recommend smallest option meeting contract; if critical assumption unresolved, one cheapest to abandon.
   • For operator_profile=business, draft owner, SMART metric, and scope fence as one confirm-with-default round.
4. Approve and hand off:
   • Self-review: placeholders, contradictions, scope, ambiguity.
   • Interactive: end with "Reply ok or corrections"; ok sets approval: approved(<operator>, <YYYY-MM-DD>) for this brief only. Brief approval alone confers NO budget, exposure, pilot, or rollout consent.
   • Autonomous or channel mode with no confirmation channel: approval: assumed-autonomous; continue.
   • Standard/Deep: save to docs/brd/brd-[slug].md when writes allowed; mint slug once. Quick: carry contract and approval in Handoff Payload.
   • Route to plan-feature; when critical value hypotheses require bounded empirical testing, route to validate-product (when selected/installed; otherwise record exact capability gap, e.g. product-management/product-experimentation not installed; opt in at a reviewed ref and sync, keep hypothesis as unvalidated assumption, and route to plan-feature).

Runtime Contract

• Use for rough feature, ops, process, or technical-direction ideas before a PRD exists for the slug. Brief approval does not grant deployment, spend, or runtime exposure authority.
• Required inputs: rough intent; missing owner, metric, or scope fence get drafted defaults, not a block.
• Return BLOCKED only when the operator rejects drafted defaults for owner, value, or scope.

Handoff Payload

• slug, operator_profile, lane, snc_tier, approval, contract (outcome, constraints, non-goals, acceptance criteria), SMART metric (Why lane), recommended/rejected options, evidence ledger (with canonical product ledger refs when present), assumptions (flagged assumed), open questions, PM handoff checklist.
• Outcome report with feature_status=requirements_ready | blocked, requirement trace seed, completed/missing evidence, decision needed, and recommended next workflow.

Blocking Questions

• Ask max 3 at a time with a recommended default and 2-3 options.

Output Template

# Brief: [Name]
lane: why | direction | both; snc_tier: low | medium | high; approval: pending | approved(<who>, <YYYY-MM-DD>) | assumed-autonomous
## Contract
Outcome; Constraints; Non-Goals; Acceptance Criteria
## Why
Business objective; SMART metric; users; problem; AS-IS to TO-BE; sponsor and validation owner
## Options And Recommendation
## Evidence
| Claim | Status |
## Standard And Deep Sections
Stakeholders; cost-benefit; delivery context; glossary; PM handoff checklist
## Approval
## Outcome Report
{schema_version: 1, run_id: "[run-id]", slug: "[slug]", workflow: brainstorm-feature, feature_status: requirements_ready, started_at: "[timestamp]", completed_at: "[timestamp]", requirement_trace: {brd_objectives: [], requirements: [], acceptance_criteria: [], srs: []}, completed_evidence: [], missing_evidence: [], decision_needed: [], recommended_next_workflow: plan-feature, cost: {source: unavailable}, agent: {identity: "[agent-identity]", model: "[model]"}}
## Open Questions
## Next Workflow
plan-feature
## Cost Report
Call `get_session_cost(workflow="brainstorm-feature")` before final handoff.