Install
$ agentstack add skill-hungv47-meta-skills-design-referral ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
Design Referral — Orchestrator
Referral — designs a viral / referral loop as loop mechanics + K-factor and cycle-time math + CAC-bounded incentive economics + a falsifiable trigger, scored by a quantitative critic. Capability metadata: [routing.yaml](routing.yaml). Methodology, philosophy, scope: [references/playbook.md](references/playbook.md) [PLAYBOOK].
Core question: "What is the actual K-factor this loop produces (invites/user × conversion-per-invite), is the cycle time short enough to compound, and does the incentive cost stay under the CAC it replaces?"
Critical Gates — read first
- Compute K, don't assert it. K = (invites sent per active user) × (conversion rate per invite). A loop without this math is a wish. Estimate each input with a labeled basis (benchmark / analog / measured); never present a guessed K as fact. K ≥ 1 = viral; 0.3–1 = a real assist that lowers blended CAC; < 0.15 with a costly incentive = kill.
- Cycle time is half the engine. Compounding depends on K AND on how fast one cycle completes (invite → new user → new user invites). A K of 0.5 at a 3-day cycle beats a K of 0.7 at a 60-day cycle. Both must be stated.
- The incentive must be CAC-bounded. Total incentive cost per acquired user (both sides of a double-sided reward) must be less than the paid CAC it displaces AND fit unit economics (LTV, margin). An incentive that costs more than buying the user is a loss; the economist agent FAILs it.
- The trigger fires at a real share-worthy moment. The loop asks for the referral right after the user got value (the activation/aha moment), not on signup. A trigger with no value-realized precondition is begging, and converts near zero.
- No fraud-blind loop, no fake scarcity, no autonomous payout. The design names the abuse vector (self-referral, fake accounts, incentive farming) and its guard. It never fabricates scarcity. It outputs a design; it never disburses a reward or moves money — payout is human-owned + out of scope.
Quality Gate — 6 dimensions
Full rubric + Pass/Fail bands: [references/rubric.md](references/rubric.md) [PROCEDURE]. Critic agent: [agents/critic.md](agents/critic.md).
- Gate: Total ≥30/42 AND every dim ≥4/7. Below either → FAIL, route to the named agent, max 2 rewrite cycles.
- Dimensions: Loop-math soundness · Cycle-time realism · Incentive economics · Trigger placement · Mechanic falsifiability · Fraud/abuse guard.
Before Starting
Per [references/_shared/before-starting-check.md](references/_shared/before-starting-check.md). Recommended context:
| Artifact | Source | Required? | |---|---|---| | research/product-context.md | research-icp | Recommended — the value moment (the share trigger) + unit economics | | research/icp-research.md | research-icp | Recommended — who the user would refer + the social context | | docs/forsvn/artifacts/marketing/design-lifecycle/*.md | design-lifecycle | Optional — the activation event the referral trigger fires after | | .forsvn/performance/*.tsv | measure-results | Optional — measured CAC the incentive must beat |
Pre-Dispatch + Mode
Run canonical Pre-Dispatch ([references/_shared/pre-dispatch-protocol.md](references/_shared/pre-dispatch-protocol.md)). Needed dimensions: product has retention/PMF (HARD precondition — a leaky product virally spreads churn) · the share-worthy moment (the trigger) · loop-type (one-sided / double-sided / milestone / collaborative-native) · unit economics (CAC-to-beat, LTV, margin) · incentive idea (or "recommend one"). Missing retention-evidence OR no share moment → BLOCK. Full Cold Start + Missing-Input Hard Blocks: [references/procedures/pre-dispatch.md](references/procedures/pre-dispatch.md).
Mode ([references/_shared/mode-resolver.md](references/_shared/mode-resolver.md)): budget: deep. --fast collapses to a single inline pass (one loop variant, K-math + economics inline, no rewrite loop). --fast does NOT skip Cold Start, Critical Gates 1-5, the retention precondition, or the no-payout gate.
Agents + Routes
4 sub-agents (loop-architect → incentive-economist + mechanic-copy → critic) across Layer 1 loop design → Layer 2 economics + mechanic → critic gate. Two routes — A (design a new loop), B (called by another skill). Full manifest + dispatch graph: [references/agent-manifest.md](references/agent-manifest.md) [PROCEDURE].
Artifact Contract
- Path:
docs/forsvn/artifacts/marketing/design-referral/[loop-type]-[date]-[slug].md - Lifecycle:
pipeline— re-run on incentive change, measured-K update, or unit-economics shift. - Frontmatter + body order + the K-factor math block schema: [
references/format-conventions.md](references/format-conventions.md) [PROCEDURE]. - Consumed by:
measure-results(the loop's measured K + payback close the loop),plan-campaign(referral as a selected channel).
Anti-Patterns
[references/anti-patterns.md](references/anti-patterns.md) [ANTI-PATTERN]. Re-read before any loop ships: asserted-K-without-math, ignoring cycle time, incentive-costs-more-than-CAC, trigger-on-signup-not-value, referral-on-a-leaky-product, fraud-blind incentive, fake scarcity, double-sided reward that breaks margin.
Durable Rules (protected)
Completion Status
- DONE — loop mechanics + K/cycle math + incentive economics + trigger + fraud guard written, critic PASS.
- DONEWITHCONCERNS — loop complete but with estimated (not measured) K inputs or thin unit economics; flagged in the artifact.
- BLOCKED — product lacks retention evidence (referral would spread churn); needs a retention/PMF decision first.
- NEEDS_CONTEXT — no unit economics available to bound the incentive; recommend
measure-results/ product-context first.
Worked Example
End-to-end double-sided loop (B2B reporting tool, K computed from labeled inputs, 3 loop variants compared on K × cycle-time × incentive-cost, CAC-bounded reward, fraud guard, critic PASS with a cycle-0 FAIL the critic caught): [references/examples/referral-walkthrough.md](references/examples/referral-walkthrough.md) [EXAMPLE].
References
references/{playbook, rubric, format-conventions, anti-patterns, agent-manifest, loop-models, incentive-economics}.mdreferences/procedures/pre-dispatch.md[PROCEDURE]references/examples/referral-walkthrough.md[EXAMPLE]agents/{loop-architect, incentive-economist, mechanic-copy, critic}.mdreferences/_shared/{pre-dispatch-protocol, before-starting-check, mode-resolver, marketing-foundations}.md
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: hungv47
- Source: hungv47/meta-skills
- License: MIT
- Homepage: https://skills.forsvn.com/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.