bounded-growth-loop
Design a small recurring product, content, SEO, analytics, or go-to-market review loop with evidence checks, durable state, explicit stop rules, and approval-gated outputs. Use before scheduling recurring growth work or turning a repeated marketing task into an agent loop.
Install
npx skills add https://github.com/markoblogo/abvx-agent-skills/tree/main/skills/bounded-growth-loop
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install markoblogo-abvx-agent-skills@llmmart
git clone https://github.com/markoblogo/abvx-agent-skills.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole markoblogo/abvx-agent-skills collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
Bounded Growth Loop
Turn a proven recurring review into a safe, useful loop. The default output is a report or draft, never an unreviewed external action.
Use When
- a content freshness, SEO, funnel, partner, activation, or market-positioning review repeats on a cadence;
- a team wants a weekly or monthly operating rhythm without a full marketing runtime;
- a recurring agent task needs state, deduplication, evidence checks, and a clear human gate.
Loop Contract
Specify all of these before scheduling:
- cadence and trigger;
- action condition, distinct from the check itself;
- one intended outcome and bounded scope;
- source and product context inputs;
- self-check and evidence threshold;
- durable state, dedupe key, cooldown, and run log;
- proposal output location;
- stop, error, and escalation rules;
- approval boundary and rollback/discard path.
Workflow
- Read
docs/product-context.mdor the repo's equivalent. - Select one narrow loop; begin report-only or proposal-first.
- Match cadence to signal speed and sample size; do not create daily noise for slow signals.
- Define the condition that authorizes a recommendation or draft.
- Retain the report, evidence, and state outside ephemeral chat.
- Run a verifier before proposing action; distinguish tracking defects, seasonality, and real changes.
- Send external copy, publishing, scheduling, spend, pricing, outreach, or product mutations through explicit approval.
- Review whether the loop is still useful; pause it when it creates noise or no one acts on its output.
Default Levels
L1 report-only: reads and summarizes with no external mutation.L2 proposal-first: creates reviewable drafts or PRs behind a human gate.L3 governed: only after durable state, verifier evidence, budget limits, run log, rollback, and controlled settlement exist.
Guardrails
- Do not auto-publish, auto-send, shift ad spend, alter pricing, or contact people.
- Do not act on sparse, stale, personally identifiable, or unverified data.
- Do not treat a dashboard movement as causation; record alternative explanations.
- Keep the loop product-specific and delete it when it no longer earns its operating cost.
Pair With
product-contextfor approved product language and boundaries;loop-readiness-reviewfor readiness classification;bounded-evaluationfor the verifier;reversible-agent-taskandsocial-publishing-gatefor proposal and publication settlement.
Provenance
Adapted from coreyhaines31/marketingskills recurring-loop structure, aligned with ABVX proposal-first execution and loop-readiness gates rather than a marketing automation runtime.
Files (abvx-agent-skills)
-
agents
-
openai.yaml 234 B
Loading…
-
-
SKILL.md 3 KB
--- name: bounded-growth-loop description: Design a small recurring product, content, SEO, analytics, or go-to-market review loop with evidence checks, durable state, explicit stop rules, and approval-gated outputs. Use before scheduling recurring growth work or turning a repeated marketing task into an agent loop. license: MIT metadata: abvx_status: experimental abvx_origin: adapted --- # Bounded Growth Loop Turn a proven recurring review into a safe, useful loop. The default output is a report or draft, never an unreviewed external action. ## Use When - a content freshness, SEO, funnel, partner, activation, or market-positioning review repeats on a cadence; - a team wants a weekly or monthly operating rhythm without a full marketing runtime; - a recurring agent task needs state, deduplication, evidence checks, and a clear human gate. ## Loop Contract Specify all of these before scheduling: - cadence and trigger; - action condition, distinct from the check itself; - one intended outcome and bounded scope; - source and product context inputs; - self-check and evidence threshold; - durable state, dedupe key, cooldown, and run log; - proposal output location; - stop, error, and escalation rules; - approval boundary and rollback/discard path. ## Workflow 1. Read `docs/product-context.md` or the repo's equivalent. 2. Select one narrow loop; begin report-only or proposal-first. 3. Match cadence to signal speed and sample size; do not create daily noise for slow signals. 4. Define the condition that authorizes a recommendation or draft. 5. Retain the report, evidence, and state outside ephemeral chat. 6. Run a verifier before proposing action; distinguish tracking defects, seasonality, and real changes. 7. Send external copy, publishing, scheduling, spend, pricing, outreach, or product mutations through explicit approval. 8. Review whether the loop is still useful; pause it when it creates noise or no one acts on its output. ## Default Levels - `L1 report-only`: reads and summarizes with no external mutation. - `L2 proposal-first`: creates reviewable drafts or PRs behind a human gate. - `L3 governed`: only after durable state, verifier evidence, budget limits, run log, rollback, and controlled settlement exist. ## Guardrails - Do not auto-publish, auto-send, shift ad spend, alter pricing, or contact people. - Do not act on sparse, stale, personally identifiable, or unverified data. - Do not treat a dashboard movement as causation; record alternative explanations. - Keep the loop product-specific and delete it when it no longer earns its operating cost. ## Pair With - `product-context` for approved product language and boundaries; - `loop-readiness-review` for readiness classification; - `bounded-evaluation` for the verifier; - `reversible-agent-task` and `social-publishing-gate` for proposal and publication settlement. ## Provenance Adapted from `coreyhaines31/marketingskills` recurring-loop structure, aligned with ABVX proposal-first execution and loop-readiness gates rather than a marketing automation runtime. -
SKILL_CARD.md 1.4 KB
Loading…
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.