Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add markoblogo/abvx-agent-skills --skill bounded-growth-loopgit clone --depth 1 https://github.com/markoblogo/abvx-agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/markoblogo/abvx-agent-skills/bounded-growth-loop)<a href="https://agentmods.dev/skills/markoblogo/abvx-agent-skills/bounded-growth-loop"><img src="https://agentmods.dev/badge/skills/markoblogo/abvx-agent-skills/bounded-growth-loop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/markoblogo/abvx-agent-skills/bounded-growth-loop"><img src="https://agentmods.dev/badge/skills/markoblogo/abvx-agent-skills/bounded-growth-loop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00056 | $0.00651 |
| Opus 5 | $0.00028 | $0.00326 |
| Sonnet 5 | $0.00011 | $0.00130 |
| Haiku 4.5 | $0.00006 | $0.00065 |
Grade A, and why
bounded-growth-loop scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.
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.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 68 lines · 56 tokens per session scan A 820797d36f52
bounded-growth-loop is a skill published in the GitHub repository markoblogo/abvx-agent-skills (16 stars, last pushed today), licensed MIT. It adds 56 tokens to every session and 651 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
cli-e2e
Write, modify, or debug Docker-based Composio CLI end-to-end tests under ts/e2e-tests/cli, including binary invocation, fixture isolation, output assertions, and package manifests. Use for CLI E2E test suites only; use cli-command for CLI source implementation.
trace
Dispatch trace subagent to investigate unknown issues — reproduces, traces, and reports root cause for fix handoff.
codex-agent
Use when you want a second-opinion review via Codex CLI, cross-verification after another agent implements changes, debugging help, or alternative implementation proposals. Requires Codex CLI to be installed and authenticated.
openlore-analyze-codebase
Run a full static OpenLore analysis and summarize architecture, call graph, refactoring issues, and duplicate code. Use when asked to analyze, map, or assess a codebase without LLM inference.
css-debug
Use this skill to diagnose CSS and frontend layout issues such as positioning, overflow clipping, Tailwind class conflicts, z-index stacking, and React rendering visibility problems.
config-validator
Validate AIWG configuration files and project setup for correctness and completeness.