Guide · Agent skill
How to use the REA Analysis Optimization agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
REA Analysis Optimization is an agent skill from the morluto/rea repository that guides an AI agent through performance work on REA's analysis providers and shared analysis workflows. It is for maintainers and contributors who are investigating a concrete latency, CPU or memory bottleneck, or designing a cache or query change. After this guide you will know what the skill changes in the agent's behaviour, how to install it, how to ask for it in plain language, and what it needs and does not do.
REA Analysis Optimizationmorluto/rea · MIT Download the skill Quick start
- 1Get the skill ZIP. Download the rea-analysis-optimization ZIP from the directory page.
- 2Install it. Unzip into ~/.claude/skills/ for personal use or into .claude/skills/ in your project, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 3Check your checkout. Confirm you have a checkout of the morluto/rea repository and can build REA's CLI or MCP tools.
- 4Make a natural request. Describe the analysis bottleneck or cache design question and the REA area it affects.
- 5Pin the baseline. Ask the agent to record the baseline revision, inputs, host and entrypoint before changing code.
- 6Measure before and after. Run the real public workflow on equivalent inputs and compare outcome, timings, CPU, peak memory and cleanup.
- 7Decide and report. Accept the change only if semantics are preserved, or record a measured decision to leave the code alone.
What is an agent skill?
An agent skill is a folder of instructions and supporting files that an AI agent can load when a task matches the skill's description. This one is defined by a SKILL.md file with a name and description, plus an agents/openai.yaml file and a reference document about a past case study.
What the skill makes the agent do
Once installed, the agent approaches REA performance work in a disciplined order. It pins the analyst's completed task and a correctness oracle that does not depend on the code being changed, then traces one public CLI or MCP task to the earliest step doing unnecessary work. It changes that step at its owner and verifies that results, coverage and Evidence stay equivalent before and after.
- Compares completed tasks before and after a change, with equivalent inputs and outputs.
- Requires an independent correctness check and careful cache key, lifetime and invalidation design.
- Reports measurements honestly, including fresh versus warm runs and failed attempts.
- Allows a measured decision to leave the code alone when a benchmark shows no real gain.
- Does not grant permission to run analysis targets, publish or merge changes.
Install the skill
You can install the skill in one of two ways. For a personal setup, download the ZIP from the directory page and unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder inside your checkout.
If you use the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project: .claude/skills/ in your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Prepare the requirements
The skill assumes you are working inside a checkout of the morluto/rea repository, because it points to the checkout's performance priorities, tool design and testing guide. You also need to be able to build and run REA's CLI or MCP tools so you can take before and after measurements.
The source does not name specific benchmark or profiling tools. You need to provide CPU time and peak memory measurement tools on your own host.
- A checkout of the morluto/rea repository
- Ability to build and run REA's CLI or MCP tools
- Benchmark and profiling tools on the host (not named by the skill)
Trigger the skill with natural requests
Describe the bottleneck or the design question in plain terms and name the REA area involved. The skill's description covers analysis bottlenecks and cache or query design, so requests on those topics are the ones it is meant to handle.
Requests about CI or test-suite speed should go to the separate rea-ci-tests-audit skill instead.
- Help me find why the Kuna string lookup is slow on large binaries and propose a fix at its owner.
- Review this cache design for a shared analysis producer: what is the key, lifetime and invalidation?
- Measure whether this change to the decoding step actually reduces peak memory for a real CLI task.
- Our test suite is slow. Can you speed it up? (Use rea-ci-tests-audit instead.)
Safety notes: scripts and network
The skill metadata states that it runs scripts and does not use the network. Measurements will execute the REA build and tools on your machine, so treat before and after runs as you would any code execution on your host.
The skill also asks the agent to stop before out-of-memory conditions and to verify cleanup of owned resources. Set CPU, memory and disk limits appropriate to your host.
- Runs scripts: yes. Benchmarks execute code locally.
- Uses network: no, according to the skill metadata.
- Permission to execute analysis targets, publish or merge still comes from you, not the skill.
Tips for good results
Give the agent a real artifact and a small or empty case, and state the baseline revision. Ask for a correctness oracle that is independent of the changed code. Expect the agent to sometimes conclude that no change is worth making.
- Use a separate checkout for experimental changes.
- Include small inputs so setup costs and regressions show up.
- Keep failed attempts in the record rather than retrying silently.
What people use it for
- Finding why an exact-address string lookup scans cached facts and proposing a narrower lookup at its owner.
- Reviewing a shared analysis producer's cache key, invalidation and lifetime before merging.
- Checking whether a projection change reduces peak memory for a real CLI task.
- Deciding that a proposed helper refactor needs only correctness tests, not a benchmark.
Frequently asked questions
Is the REA Analysis Optimization skill free and under what licence?
The skill is listed under the MIT licence. The source does not describe any pricing, so check the directory page for terms.
What do I need to install the rea-analysis-optimization skill?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. You can also upload it under Settings → Capabilities → Skills in the Claude apps.
Does the skill run code on my machine?
The metadata says it runs scripts. Performance measurements execute REA's build and tools locally, so review what you are running before you approve it.
Does this skill work for CI or test-suite speed?
No. The skill assigns CI and test-suite optimization to the separate rea-ci-tests-audit skill.
Can I use this skill just to analyse a binary with REA?
Not really. It is about changing REA's implementation, and the source warns it can lead to risky edits if applied carelessly. General analysis users do not need it.
What tools do I need for benchmarking?
The skill refers to measuring CPU time and peak memory but does not name specific tools, so you must supply your own profiling and benchmark tools on the host.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you maintain or contribute to REA's analysis providers and need a structured way to find, change and verify a real bottleneck. Pick something else for routine refactoring, CI or test-suite speed, or simply analysing a binary, since the skill is built for changing REA's implementation carefully.
Download the skill