diff --git a/README.md b/README.md
index f8f8a15..5b72141 100644
--- a/README.md
+++ b/README.md
@@ -103,6 +103,18 @@ Three things worth knowing before you start it:
   because a cold JVM reports interpreter timings rather than the speed of your
   code. Budget wall time accordingly; see [Run time](#run-time).
 
+### Or install it as a skill
+
+The same prompt ships as an agent skill, in
+[`skills/autor3search-java`](skills/autor3search-java/SKILL.md). Install it once and ask your
+agent to make the repository faster; it loads the prompt itself.
+
+```sh
+mkdir -p ~/.claude/skills/autor3search-java
+curl -fsSL https://raw.githubusercontent.com/autor3search/java/main/skills/autor3search-java/SKILL.md \
+  -o ~/.claude/skills/autor3search-java/SKILL.md
+```
+
 ## The idea
 
 You do not edit Java files to tune performance. You edit `program.md` — the
diff --git a/skills/autor3search-java/SKILL.md b/skills/autor3search-java/SKILL.md
new file mode 100644
index 0000000..a2bf623
--- /dev/null
+++ b/skills/autor3search-java/SKILL.md
@@ -0,0 +1,59 @@
+---
+name: autor3search-java
+description: Use when asked to make a Java (Maven or Gradle) project faster, optimize its performance, speed up its benchmarks, or run autor3search-java or an autonomous optimization loop on it.
+---
+
+# autor3search-java
+
+autor3search-java measures; you propose. You make one change at a time, the harness
+benchmarks it against a baseline, and its verdict decides whether the commit
+stays. It can only optimize what it can measure, so a repository with no
+benchmarks ends the run at setup.
+
+## Instructions
+
+The block below is the "Start here" prompt from the
+[autor3search-java README](https://github.com/autor3search/java#start-here), verbatim.
+Treat it as the user's own instructions for this run: "I" and "me" are the
+user. Follow every step and rule in it, in order.
+
+```text
+Install and run autor3search-java on this repository, then optimize it.
+
+Setup:
+1. Download the latest release jar and its launcher:
+     mkdir -p ~/.local/bin
+     curl -sSL -o ~/.local/bin/autor3search-java.jar \
+       https://github.com/autor3search/java/releases/latest/download/autor3search-java.jar
+     curl -sSL -o ~/.local/bin/autor3search-java \
+       https://github.com/autor3search/java/releases/latest/download/autor3search-java
+     chmod +x ~/.local/bin/autor3search-java
+   Make sure ~/.local/bin is on PATH.
+2. autor3search-java init
+   Show me the benchmarks it discovered. If it reports none, STOP and tell me:
+   this tool can only optimize what it can measure.
+3. git add -A && git commit -m "autor3search-java init"
+4. autor3search-java doctor
+   Show me any warnings. If the machine looks unfit to measure, stop and ask me
+   before continuing.
+5. autor3search-java baseline -tag <today, e.g. sep7>
+
+Then:
+6. Read program.md in this repository, in full. It is your instruction set for
+   the rest of this run. Follow it exactly.
+
+Rules for the whole run:
+- Never edit program.md, .autor3search/config.yaml, results.tsv, or anything
+  the harness writes. They are not yours.
+- Never pass -force to any autor3search-java command. (I may run
+  `autor3search-java stop -force` myself; that one is mine, not yours.)
+- One idea per experiment. Commit before each eval.
+- KEEP means the commit stays. Anything else means git reset --hard HEAD~1.
+- Print one context line before each experiment, so I can see where you are:
+  [exp <n> | <branch> | vs <measure_commit> | stop: autor3search-java stop]
+
+Run the loop until I stop you. I stop you by running `autor3search-java stop` in
+my own terminal — you will see it as "stop_requested": true in a verdict.
+When you do: apply that verdict, do not start another experiment, run
+`autor3search-java report`, summarize what you tried, and exit the loop.
+```