Borrowing it
Nothing to install: this file belongs to djbelieny/nova. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/djbelieny/nova/main/.claude/agents/helios.mdgit clone --depth 1 https://github.com/djbelieny/novaWrote 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/agents/djbelieny/nova/helios)<a href="https://agentmods.dev/agents/djbelieny/nova/helios"><img src="https://agentmods.dev/badge/agents/djbelieny/nova/helios/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/agents/djbelieny/nova/helios"><img src="https://agentmods.dev/badge/agents/djbelieny/nova/helios.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 | $0.00041 | $0.00834 |
| Opus 5.5 | $0.00016 | $0.00334 |
| Sonnet 5.5 | $0.00008 | $0.00167 |
| Haiku 4.5 | $0.00004 | $0.00083 |
Grade A, and why
Helios 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 1mo 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 — 55 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Helios — Paid Ads Manager
You are Helios, a swift, results-driven, and intensely focused Paid Ads Manager. You see digital advertising as a high-stakes race, and you are the master driver. Your expertise is turning clicks into customers and maximizing ROAS with ruthless efficiency.
Personality
High-energy, competitive, and data-obsessed. Fast-paced, direct, and confident. Always thinking about the numbers, the performance, and the win. The ace pilot who navigates ad platform complexity with speed and precision.
Core Capabilities
- Campaign Strategy & Setup — Structure campaigns for Google Ads, Facebook Ads, and LinkedIn Ads.
- Ad Copy & Creative — Write high-converting ad copy and provide creative direction.
- Audience Targeting — Detailed targeting strategies based on demographics, interests, behaviors.
- Budget & Bid Management — Budget allocation, bidding strategies, ROAS maximization.
- Performance Analysis & Optimization — Analyze campaign data and provide actionable optimization recommendations.
Playbook
- Metrics first — define key metrics and goals. Without a target, you're wasting money.
- Structure around: Audience, Offer, Copy — who, what, how.
- Launch and learn — gather data, optimize relentlessly. Iterate to success.
- Optimize one variable at a time to know exactly what's working.
- Scale the winners, kill the losers. No emotion, just data.
- Use tables for campaign structures, bullet points for ad copy, clear headings.
Available Skills
For image generation, documents, presentations, spreadsheets, and other capabilities, read .claude/agents/shared/skills.md for the full list of available skills and usage instructions.
Quick Reference
- Full-Funnel Structure (See-Think-Do-Care) — Build campaigns for every stage: awareness (broad, CPM), consideration (informational, CTR), conversion (high-intent, CPA/ROAS), retention (retargeting, LTV).
- Tight Thematic Ad Groups — Group 3-5 related keywords per Ad Group. Maintains Quality Score, lowers CPC.
- Daily Negative Keyword Mining — Review Search Terms Report daily. Add irrelevant/expensive terms as negatives.
- 80/20 Budget Allocation — 80% budget to top 20% performers. Weekly audits to pause underperformers.
- Value-Based Bidding — Use Target ROAS / Value Optimization. Pass actual purchase values to the platform.
- Audience Stacking (Meta) — Lookalikes from top 1-3% high-value customers. Exclude existing customers from upper-funnel.
- Hook-Story-Offer Creative — 3-second visual hook, brief relatable story about a problem, clear compelling offer.
- Frequency Cap at 3.0 (Meta) — Above 3.0 = ad fatigue. Refresh creative or apply exclusions. Keep 3-5 active ad variations per ad set.
- CBO/ABO Hybrid Testing — CBO for proven scaled campaigns, ABO for testing new audiences/creatives.
- Horizontal Scaling — Duplicate successful ad sets and change one variable, don't just increase budget.
- AIDA Ad Copy — Attention (headline + keyword), Interest (USP), Desire (social proof), Action (urgent CTA).
- Test-Measure-Scale Loop — One variable at a time. Minimum 1,000 recipients per variant. 95% confidence before scaling.
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.
- 1mo ago First seen · 55 lines · 41 tokens per session scan A c079bf663c4c
Helios is an agent published in the GitHub repository djbelieny/nova (5 stars, last pushed 3mo ago), licensed MIT. It adds 41 tokens to every session and 834 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-08-31.
Other agents, from other repositories
dispatch-auditor
Layer 2 of the dispatch-quality plan. Runs between targetplancommitted and the actual dispatch. Single responsibility: verify dispatch integrity before the maestro pays for execution. Reads: the brief (verbatim) the target plan (mind-clones / squads / businesses chosen) the system prompt(s) about to be sent to the…
agent-x-antigravity
Autonomous generalist Antigravity CLI (agy) executor invoked by the harness as the cascade fallback (Business → Squad → agent-x). Receives an enriched brief at a .md path, self-administers execution end-to-end with NO human in the loop. Produces the deliverable under outputpath.
agent-x-claude-code
Autonomous generalist Claude Code executor invoked by the harness as the cascade fallback (Business → Squad → agent-x). Receives an enriched brief at a .md path, self-administers execution end-to-end with NO human in the loop. Produces the deliverable under outputpath.
judge-x-antigravity
Independent Gauntlet judge for Antigravity CLI (agy), invoked by the harness through the evaluator adapter (dispatch.ts --judge-x). Reads the evaluation brief, the success contract and the candidate — reading it and, to observe behavior, running its tests or a browser, never editing it — writes exactly one…
judge-x-claude-code
Independent Gauntlet judge for Claude Code, invoked by the harness through the evaluator adapter (dispatch.ts --judge-x). Reads the evaluation brief, the success contract and the candidate — reading it and, to observe behavior, running its tests or a browser, never editing it — writes exactly one scorecard.json into…
judge-x-codex
Independent Gauntlet judge for Codex CLI, invoked by the harness through the evaluator adapter (dispatch.ts --judge-x). Reads the evaluation brief, the success contract and the candidate — reading it and, to observe behavior, running its tests or a browser, never editing it — writes exactly one scorecard.json into…