Strategy — Step 2 of 4. Brainstorms strategic solutions to a confirmed problem, surfaces unconventional tactics, and ranks them with evidence-backed scoring against explicit trade-offs.
Core Question: "What's the highest-impact thing we can do about this?"
[Read references/playbook.md [PLAYBOOK] for why this skill exists, the anti-generic test, forced-ranking ceiling, ≤3 cut-line philosophy, churn special case, unconventional-scan rationale, and when NOT to use.]
Critical Gates — Read First
- Every initiative hypothesis MUST name the root cause. Generic growth ideas generate low-impact, unfocused initiatives. If the hypothesis "because" clause doesn't reference the confirmed root cause / problem, the initiative is untethered.
- Force-rank BEFORE scoring — ranking prevents "everything is a 6." The forced ranking sets the ceiling for ICE scores. If you ranked it #1, its ICE should be highest.
- Anti-generic test is mandatory — "would this help ANY company?" Delete the root cause reference from an initiative. If it still makes sense for any company, it's generic. Rewrite.
- ≤3 initiatives above cut line — force the constraint. More than 3 active initiatives means none get full attention. Surface trade-offs honestly; parked initiatives have their turn after the current batch ships.
Before Starting
Apply the before-starting-check [PLAYBOOK]:
- Mode resolution per
references/_shared/mode-resolver.md[PROCEDURE]. Skill isbudget: deep;--fastforces Route B (skip unconventional-agent) with critic gate collapsed to single pass. Hard gate (Critical Gates above) STILL enforced under--fast— safety gates supersede mode-resolver downgrade. Session execution profile (single-vs-multi): inherit perreferences/_shared/execution-policy.md. - Read
implementation-roadmap/canonical-paths.mdif present — verify output path matches canonical inventory (docs/forsvn/artifacts/meta/sketches/prioritize-*.md) and Out-of-Scope path (docs/forsvn/artifacts/meta/out-of-scope/). - Read
.forsvn/index/manifest.json— find the matchingdiagnose-*.md(required) and priorprioritize-*.md(if any). Check freshness (>30 days surfaces a warning). - Run Pre-Dispatch per
references/procedures/pre-dispatch.md[PROCEDURE] — hard-gate enforcement, read order, Cold/Warm Start prompts, staleness check, constraint interview, Out-of-Scope persistence on write all there.
Artifact Contract
- Path:
docs/forsvn/artifacts/meta-prioritize-<YYYY-MM-DD>-<slug>.md(flat v2; one per root cause; re-run renames prior with.v[N]) - Lifecycle:
sketch· Review surface:none(decision_state: not_required) - Frontmatter:
skill,version,date,status,stack=meta,review_surface=none - Body sections (cross-stack contract): Phase 1 Initiatives (5-10 + 2-4 unconventional) · Phase 2 (Forced Ranking, ICE Scoring, Decisions) · Cut line · Next Step
- Side effects on PASS: one file per Kill written to
docs/forsvn/artifacts/meta/out-of-scope/[name].md; experience write-back is NONE (initiatives are project-specific, not stable profile state)
Full schemas, consumer map, and cross-stack load-bearing field list: references/procedures/artifact-contract.md [PROCEDURE].
Agent Manifest
| # | Agent | Layer | Focus |
|---|---|---|---|
| 1 | research-agent | L1 | Root cause validation, case studies via WebSearch, constraint summary |
| 2 | initiative-generator-agent | L1.5 ‖ | 5-10 standard initiatives with Hypothesis/Mechanic/Target Metric/Anti-generic check |
| 3 | unconventional-agent | L1.5 ‖ | 2-4 asymmetric/non-obvious tactics with risk assessments |
| 4 | ranking-agent | L2 seq | Strict 1-through-N forced ranking with reasoning per rank |
| 5 | ice-scoring-agent | L2 seq | Evidence-backed ICE scores + differentiation check (no >2 sharing a total) |
| 6 | cut-line-agent | L2 seq | Proceed/Park/Kill decisions, capacity assessment, Proceed validation |
| 7 | critic-agent | L2 final | 9-point quality gate (rubric + failure routing in agent file). Max 2 rewrite cycles |
Routing + Dispatch
Two routes; chosen at Pre-Dispatch and echoed in the Warm Start confirmation (operator can override):
| Route | When |
|---|---|
| A — Full Analysis (default) | Any non-trivial solution design |
| B — Quick Design | User has candidate approaches; speed > breadth; OR --fast flag |
Entry mode is orthogonal to route: root-cause (default, diagnose-gated) vs ideation (input is a candidate set — discover shortlist, debate output, or a "generate ideas and rank them" ask; the diagnose gate is replaced by a stated ranking anchor and Phase 1 may expand at volume). Resolution + mechanics: references/procedures/pre-dispatch.md § Entry-mode resolution.
Mechanics (route graphs, Layer 1/1.5/2 spawn details, merge step + user feedback gate, critic FAIL routing, single-agent fallback, post-write side effects, Out-of-Scope Persistence file format, chain position, skill deference) live in references/procedures/dispatch-mechanics.md [PROCEDURE]. Load at Layer 1 dispatch entry.
Anti-Patterns
Critic-load reference: references/anti-patterns.md [ANTI-PATTERN]. Re-read before any output ships — 8-pattern catalog (mapped onto the critic's 9-point quality gate in agents/critic-agent.md) plus 2 cross-cutting failures (cross-stack contract drift + out-of-scope persistence skipped).
Durable Rules (protected)
Completion Status
Every run ends with explicit status:
- DONE — initiatives generated, ICE-scored, ranked, cut-line drawn (≤3 above), kill criteria attached, critic PASS
- DONE_WITH_CONCERNS — ranking complete but with sizing/impact uncertainty flagged at item level (e.g., effort estimates speculative, ICE inputs from interview not data); OR critic loop cap reached with surfaceable gate failures (pinned at top as Known Issues)
- BLOCKED — root-cause mode:
id:diagnoseunresolvable (no diagnose artifact in the manifest) AND no other root-cause source available; STOP gate per Critical Gate semantics — recommenddiagnosefirst (hard-gated, no INTERVIEW substitute). Ideation mode: no ranking anchor obtainable - NEEDS_CONTEXT — diagnose available but
id:product-contextunresolvable for impact estimation; recommendresearch-icp
Next Step
After cut line is drawn: feed Proceed initiatives forward — plan-funnel (numeric targets per initiative), plan-campaign (campaign briefs for Proceed items), architect-system (technical scoping per Mechanic). Re-run prioritize when new root cause surfaces or after a batch ships.
References
references/playbook.md— why this exists, anti-generic test, ≤3 cut-line philosophy, churn special case, when NOT to usereferences/procedures/{pre-dispatch, dispatch-mechanics, artifact-contract}.md— read-order + hard gates · route graphs + side effects · full contract + consumer mapreferences/_shared/{before-starting-check, mode-resolver, pre-dispatch-protocol, hypothesis-framework, artifact-contract-template}.mdreferences/format-conventions.md— Phase 1 format, ICE/Decisions table schemas, Cut line, Next Step, Out-of-Scope file formatreferences/anti-patterns.md— 8 named anti-patterns + cross-stack drift + out-of-scope persistence skippedreferences/examples/prioritize-walkthrough.md— full Route A walkthrough on a 2-root-cause acquisition + activation case- Data catalogs:
initiative-types.md·ice-scoring-rubric.md·initiative-planning.md·churn-{playbook, cancel-flow-templates, health-score-guide}.md research-skills/CLAUDE.md— stack-level conventions