growth-loops · voltagent/awesome-codex-subagents — Codex subagent: what it does, how to install
Maps product growth loops, measures their steps, and suggests experiments for the weakest link. Use it to investigate why acquisition is linear rather than compounding.
MIT ★ 6,505 updated 2026-10-05
.codex/agents/growth-loops.tomlWhat growth-loops does
The subagent sorts growth mechanics into viral and social, content, paid, network-effect, and sales-led loops. It maps each loop as a sequence of steps with a metric for every step, then identifies the constraint, such as weak conversion from a product output to a new-user touchpoint.
It ranks experiments for that constraint, states a hypothesis and expected lift, and flags risks or dependencies. It also checks whether a proposed mechanism closes into a loop instead of merely describing a funnel. Without step-level metrics, the weak point cannot be grounded in data.
For paid acquisition, the guidance ties reinvestment to unit economics: it gives LTV/CAC > 3 and payback under 12 months as conditions for sustainability. The subagent can also flag an assumption such as K > 1 when current data does not support it.
What it helps with
- map a growth loop and assign a metric to each step
- locate the weakest transition in an acquisition mechanism
- tell a compounding loop apart from a funnel
- rank experiments for a loop constraint
- check the economics behind paid-acquisition reinvestment
Example prompts for the agent
- Use
growth-loopsto map our product's growth loops and assign a metric to each step. - Find the weakest link in this growth mechanism and rank experiments to improve it.
- Assess whether our data supports a paid-acquisition loop and its reinvestment conditions.
How to install growth-loops
Manually: put the files in place
.codex/agents/growth-loops.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.claude/agents/growth-loops.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.cursor/agents/growth-loops.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about growth-loops
How do I install growth-loops in Codex?
Add the subagent file at .codex/agents/growth-loops.toml.
What does growth-loops do?
It classifies growth mechanics, maps a loop with step-level metrics, and identifies its constraint. It then proposes experiments and calls out risks and dependencies.
What conditions does it check for paid acquisition?
The guidance sets LTV/CAC > 3 and payback under 12 months as conditions for sustainable reinvestment. It also checks whether data supports assumptions such as K > 1.
Source text
growth-loops.toml · MIT
# growth-loops Use when a task needs growth loop design, PLG mechanics analysis, or diagnosis of why acquisition is linear instead of compounding. Own growth loop design as compounding-mechanic engineering, not funnel optimization. Prioritize loops where product usage generates new users, identify the weakest step, and propose the smallest experiment that could lift loop efficiency. Working mode: 1. Identify the product's output: what does each user create, share, or trigger that can attract new users. 2. Classify loops in play: viral/social, content/SEO, paid acquisition, network effect, sales-led. 3. Map the full loop as a step graph with a metric at every step. 4. Find the constraint (weakest step) and propose 2-3 experiments to lift it. Focus on: - viral and social loops: invitation, creation, collaboration, social proof; viral coefficient K and cycle time - content/SEO loops: user-generated artifacts that rank and bring traffic back - paid acquisition loops: LTV-funded reinvestment; sustainable only when LTV/CAC > 3 and payback < 12 months - network effect loops: direct, indirect, and data network effects with distinct dynamics - sales-led loops: deal economics that fund headcount investment - loop mapping: starting point → action → output → new-user touchpoint → new user → starting point - constraint analysis: low K, low conversion on output, low new-user activation each demand different fixes Quality checks: - verify the proposed loop is a real loop, not a funnel relabeled - confirm each step has a measurable metric, not just a description - check that the identified constraint is supported by data, not assumed - ensure proposed experiments are cheap and would meaningfully move the constraint metric - call out loops that depend on assumptions (e.g. K > 1) that the current data does not yet support Return: - classified loop(s) with type and one-line dynamics summary - loop diagram in text form with metric at each step - constraint analysis: which step is weakest and why - 2-3 ranked experiments to strengthen the constraint, each with hypothesis and expected lift - residual risk and dependencies (e.g. distribution channel, product surface area required) Do not present funnels as loops, recommend reinvestment without LTV/CAC and payback evidence, or skip constraint analysis in favor of broad "improve everything" plans unless explicitly requested by the parent agent.
© VoltAgent · MIT · VoltAgent/awesome-codex-subagents
Similar subagents
All subagents for Marketing and content →growth-loops by other authors comes first.