# skillz Load personal Pi skills only where they make sense. Use it to keep machine-specific, project-specific, or credential-dependent skills out of sessions where they cannot help. ## Set up a conditional skill Place a normal Pi skill under `~/.pi/agent/skillz/` instead of `~/.pi/agent/skills/`. If you changed Pi's agent directory, use its `skillz/` subdirectory instead. Add at least one condition to its frontmatter. ```yaml --- name: jj-workflow description: Use in a Jujutsu working directory on this machine. os: [lnx, mac] host: desktop path: .jj --- ``` This skill loads only on Linux or macOS, on host `desktop`, inside a Pi working directory containing `.jj`. Use `os` with `lnx`, `win32`, or `mac`. Use `host` for an exact hostname, `path` for a marker below Pi's current directory, `which` for an executable on `PATH`, or `env` for a nonempty environment variable. A list means any value matches. Use `{ all: [...] }` when every listed value must match. Different condition types must all match. ## Use it in Pi Nothing to invoke after setup. Eligible skills become Pi's normal `/skill:` commands when Pi starts or reloads. Run `/skillz` to see which conditional skills loaded or why a discovered candidate was excluded. The report shows environment variable names and whether they are present, never their values. Invalid conditions or duplicate skill names produce warnings and do not load that candidate. Install through [Installation](../../README.md#install). ## Reference ### Commands | Command | Purpose | | --- | --- | | `/skillz` | Inspect conditional skills from the last discovery |