How Federico Viticci Used Opus 5.5 to Trim His Open Minis Prompts
Built by Federico Viticci · · Source: X
Result: Opus 5.5 worked for 12 hours, used 9% of their Claude subscription, and cut their prompts to half the size.
What they built
Federico Viticci used Opus 5.5 to clean up and optimize their agent setup in Open Minis for iOS. They didn't write instructions from scratch. They used a prompt that Eric Zakariasson shared, which Zakariasson says is based on what Cursor has learned about making an agent harness use fewer tokens.
How it works
The prompt tells the model to lower the token cost per completed task without making the agent worse at its job. It works in a set order: map the harness and measure a baseline, rank the opportunities, make the safe changes directly, and put the rest behind flags or in proposals. It covers system prompt trimming, offloading rarely used tools, cache layout, how tool results are handled, compaction, and subagents. One of its main ideas is that capable models need plain descriptions instead of long lists of commands and warnings.
Viticci doesn't say which parts of the prompt applied to their setup or what exactly changed, beyond their prompts getting shorter.
The result
According to Viticci, Opus 5.5 ran for 12 hours, used 9% of their Claude subscription, and cut their various prompts to half their size. They called it a very good cleanup.
How to build something similar
- Collect every prompt, instruction file and tool definition your agent setup uses, so the model can see all of it at once.
- Give a capable model a clear goal, like spending fewer tokens per task with no drop in quality, and ask it to measure a baseline before changing anything.
- Have it sort instructions into keep, rewrite and delete, and send risky changes to you as proposals instead of applying them.
- Run your usual tasks on the trimmed setup and compare results before you drop the old version.
The original post
I used this prompt to have Opus 5.5 optimize my setup in Open Minis for iOS. It worked for 12 hours, used 9% of my Claude sub, and slimmed my various prompts to half the size. Very good cleanup, and crazy efficiency in Opus 5.5 right now. [Quoted post from @ericzakariasson]: here's a prompt to improve your agent harness based on what we've learned at cursor. enjoy # Improve this agent harness's token efficiency You're working on an LLM agent harness: the system prompt, tool definitions, request assembly, context caching, compaction, and retrieval, and how work is split across agents. Make the agent's runs cheaper without making it worse at its job. - Objective: lower price-weighted token cost per completed task. - Constraint: no measurable drop in task quality. Measure per task, not per request. Every turn resends the prefix (tools, instructions, setup, and the conversation so far), so a change that shrinks each request but adds turns can cost more. Weight tokens by billing type: output, uncached input, and cached input are priced very differently. Work in this order: map the harness and measure the baseline, rank the opportunities, make the changes that are safe to make directly, put the rest behind flags or in proposals, then report. Figures below come from one team's production coding agent and its multi-agent experiments. Use them to gauge magnitude, not as targets. One round of these changes (prompt trimming, tool offloading, cache layout, sparse line numbers, subagent tuning) cut that team's overall token cost about 7% with no loss in quality. The larger percentages apply only to the part of the request each change touched. ## Principles 1. Change what the harness sends, not how hard the model tries. Don't ask the model to conserve tokens. A harness that told its model to "take care to preserve tokens and not be wasteful" found it grew reluctant to take on ambitious tasks and sometimes quit, saying it wasn't supposed to waste tokens. 2. Capable models need definitions, not commands. Lists of "DO NOT", "You must", and "Important", and guards against older models' habits, can usually be replaced with plain descriptions of what each tool does. One team cut about two-thirds of its system prompt this way, and the shorter prompt worked across model families. Instruct only on what the model can't know (the product, the environment, the user's processes) and on quirks you've seen in transcripts. 3. Static context is for what most turns need. Everything else should be discoverable when needed. Less up-front context also means less confusing or contradictory information. 4. Expect removals to win. Guardrails written for weaker models, coordination steps that became bottlenecks, and prompting for behavior the model now does on its own all cost tokens. 5. Real usage decides. Evals are a fast proxy, but they skew toward hard problems and miss the real mix of requests. ## 1. Map the harness and measure the baseline Find: - Where requests are assembled, the system prompt, and tool schemas. If a framework or SDK builds requests, find its hooks for message order, cache control, and tool loading. - How tool results are formatted, and how history is kept, trimmed, or summarized. - How subagents or parallel agents are spawned, if any. - Which models and provider APIs are used. From the provider's docs, get the prompt caching behavior (automatic or explicit breakpoints, TTL, minimum cacheable length) and the prices for output, uncached input, and cached input. - Existing logging, token accounting, and evals. If the harness doesn't record per-request token usage by billing type and cache hits, add that first. Everything later depends on it. Then render a few real requests (from logs, or by running representative tasks) and count tokens per section with the model's tokenizer or the API's usage fields. Produce: - Cost share by source × billing type. Sources: system prompt, tool definitions, skill/rule/integration descriptions,
Want to build something like this?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week.
Join Shipping Skool — FreeMore use cases
- How Andrej Karpathy Built HackerTube with Claude Code and Opus 5.5Developer
- How Miguel Ríos Berríos Built Recall to Search Past Agent SessionsDeveloper
- How Adonias Built a CRM in Claude Code with Opus 5.5 and 6 SubagentsDeveloper
- How Kieran Klaassen Built a Webcam Heart Rate App with Gemini 4 ArgonDeveloper