plan-budget

Plan Budget — Orchestrator

Allocates a fixed paid budget across already-selected channels: CAC/LTV-aware mix, marginal-return reasoning over diminishing-returns curves, min-viable-spend floors, and explicit reallocation triggers. Multi-agent allocate → constraint-check → critic pipeline. Capability metadata lives in routing.yaml. Agent table + 3 routes + 7-dim rubric: references/agent-manifest.md. Methodology, marginal-return model, scope boundary, history: references/playbook.md.

Core question: Does every dollar sit where its next dollar earns the most — and is every CAC/LTV number sourced, not guessed?

Critical Gates — load first

1. Never fabricate CAC or LTV. Every channel's CAC/LTV is sourced (measure-results .forsvn/performance/*.tsv, operator-supplied + labeled, or a cited external benchmark) or it does not enter the math. A channel with no sourced CAC and no measure-results history → it is a hypothesis lane (capped test budget, labeled cac: unsourced), never an allocation backed by a number. Zero sourced inputs → NEEDS_CONTEXT (recommend running ads + measure-results first); do NOT invent a CAC.
2. Hard-block Pre-Dispatch inputs. Total budget · objective (acquisition / retention / mixed) · ≥1 channel with a sourced CAC/LTV · time horizon · the selected channel set (from plan-campaign). Missing any → BLOCK and ask one question. Missing the channel set → defer to plan-campaign first (this skill allocates across selected channels; it does not select them).
3. Marginal return, not average. Allocation is driven by each channel's next-dollar return (marginal CAC at the proposed spend on its diminishing-returns curve), never by average CAC alone. A channel cheap on average but already past its saturation knee gets less, not more. Even-split-without-reason is an automatic critic FAIL.
4. Every channel respects its floor and its §0 veto. Below min-viable monthly spend a channel cannot exit the learning phase — allocate the floor or allocate zero, never a sub-floor sliver. Never allocate to a channel whose plan-campaign launch-channel pack §0 veto fires for this ICP/stage without an explicit override + reason.
5. No autonomous spend — ever. Output is a plan. It never moves money, never connects to an ad manager, never toggles a campaign. The publish/spend gate is hard-stopped for a human (gate_class: review; spend is human-owned).

Quality Gate — 7 dimensions

Full rubric + Pass/Fail bands: references/rubric.md. Critic logic + rewrite routing: agents/critic.md.

• Gate: Total ≥49/70 AND every dim ≥6. Total 49-55 with all dims ≥6 = PASS as DONE_WITH_CONCERNS. Any dim <6 = FAIL.
• The seven dims: Input integrity (sourced CAC/LTV) · Marginal-return soundness · Diminishing-returns respect · Floor + veto compliance · Concentration balance · Reallocation triggers · Objective fit. Max 2 rewrite cycles; cycle 3 auto-surfaces with the scorecard.

Before Starting

Apply references/_shared/before-starting-check.md. Recommended context:

Artifact	Source	Required?
.forsvn/performance/*.tsv	measure-results	Strongly recommended — sourced CAC per channel; the substrate that keeps Gate 1 honest
docs/forsvn/artifacts/marketing/campaign-plan.md	plan-campaign	Recommended — the selected channel set + per-channel network + §0 vetoes
research/icp-research.md	research-icp	Recommended — LTV inputs (retention shape, ACV, payback tolerance)
research/product-context.md	research-icp	Recommended — pricing / margin for LTV; objective framing
docs/forsvn/experience/{business,product}.md	(any skill)	Optional — persisted CACs / prior allocations

Pre-Dispatch + Mode

Run canonical Pre-Dispatch (references/_shared/pre-dispatch-protocol.md). Needed dimensions: total budget (monthly) · objective (acquisition / retention / mixed) · time horizon (allocation period in months) · selected channel set (from campaign-plan or operator) · per-channel CAC (sourced — value + source label) · per-channel LTV or payback ceiling · per-channel current spend + observed saturation signal (optional but sharpens the curve) · constraints (channels off-limits, max-concentration cap, min-viable floors if non-default).

Cold Start (7-question) + Missing-Input Hard Blocks (5 conditions) + the sourced-vs-unsourced CAC classification: references/procedures/pre-dispatch.md.

Mode (references/_shared/mode-resolver.md): budget: deep. --fast collapses the allocator's curve-fitting to a coarser three-band (under/at/over saturation) pass and skips the second constraint sweep. --fast does NOT skip Cold Start, Critical Gates 1-5, Missing-Input Hard Blocks, the CAC-fabrication ban, or the critic floor. Session execution profile (single-vs-multi): inherit per references/_shared/execution-policy.md.

Routes

Three routes — A (quick: ≤3 channels, operator-supplied CACs), B (full: campaign-plan + measure-results grounded), C (called by another skill — reuses a cached allocation or runs Route B if stale). Full dispatch graphs: references/agent-manifest.md.

Artifact Contract

• Path: docs/forsvn/artifacts/marketing/plan-budget/[date]-[slug].md (allocation table + marginal-return rationale + reallocation triggers). Single artifact per run; overwrite on re-run unless version preservation via [date]-[slug].v[N].md.
• Lifecycle: pipeline — re-run on a CAC update, a budget change, a new channel selection, or a fired reallocation trigger.
• Frontmatter: skill, version, date, status, stack: marketing, review_surface, the v3 instruction core (id, type, keywords) + domain fields total_budget, objective, horizon_months, channels, cac_sourcing (sourced | mixed | hypothesis-only), critic_total. Per-field rules: references/format-conventions.md.
• Consumed by: plan-campaign (per-channel execution weight), write-ad (per-network spend context), measure-results (closes the loop — did the allocation's predicted CACs hold?).
• Cross-stack contract: schema changes require atomic update of format-conventions.md § "Frontmatter field order" + § "Field values" — never silently drift.

Anti-Patterns

Read references/anti-patterns.md before any allocation ships. Section 1 — Allocation (even-split-without-reason · sunk-cost doubling-down · average-CAC instead of marginal · over-concentration · ignoring channel saturation · sub-floor slivers · allocating into a §0-veto channel). Section 2 — Input integrity (fabricated CAC · stale CAC shipped silently · unsourced external benchmark dressed as own data · LTV inflated to justify a channel). Section 3 — Process (no reallocation trigger · objective-mismatch allocation · collapsing into plan-campaign). Each row carries a falsifiable detection rule the critic applies.

Durable Rules (protected)

Completion Status

• DONE — allocation across the selected channels with marginal-return rationale + reallocation triggers, critic PASS, every CAC/LTV sourced.
• DONE_WITH_CONCERNS — delivered with a flagged risk (one hypothesis lane on an unsourced CAC, thin saturation data, total 49-55 with all dims ≥6) named in the artifact.
• BLOCKED — total budget, objective, time horizon, or the selected channel set missing and unresolvable.
• NEEDS_CONTEXT — zero sourced CAC inputs and no measure-results history; recommend running ads + measure-results first, OR plan-campaign to select channels. Never fabricate a CAC to proceed.

Execution

This skill never spends. The allocation is a brief for a human to enter into the ad managers themselves. No Assisted/Direct execution fork — there is no render or publish step. The reallocation triggers tell the operator when to re-invoke (a channel's measured CAC breaches its trigger, the budget changes, or plan-campaign reselects channels).

Worked Example

End-to-end Route B (a real $6,000/mo split across 4 channels — Meta / Google Search / LinkedIn / newsletter — with marginal-CAC curves, two channels at their saturation knee, one hypothesis lane on an unsourced CAC, a min-viable floor binding on LinkedIn, per-channel reallocation triggers, constraint-checker bounce on an over-concentration cap, critic PASS 60/70) + a cycle-2 FAIL (fabricated CAC) trace: references/examples/plan-budget-walkthrough.md.