Agent Files
How Hyprlayer installs its skills and sub-agents for Claude Code and Codex
Hyprlayer installs the same skills and sub-agents for both Claude Code and Codex. There is no tool to select. The first hyprlayer command you run downloads the release’s hyprlayer-assets-claude-<version>.tar.gz and hyprlayer-assets-codex-<version>.tar.gz, verifies their SHA256 digests, and links the files into place.
OpenCode reuses these bundles: Claude models use the Claude bundle, and every other model uses the Codex bundle.
Where the Files Land
Section titled “Where the Files Land”Bundle bytes live in ~/.config/hyprlayer/agents/<version>/{claude,codex}. Each harness gets per-entry symlinks (directory junctions on Windows) into that generation, so your own skills and agents in the same directories stay untouched.
| Harness | Skills | Agents |
|---|---|---|
| Claude Code | ~/.claude/skills/ |
~/.claude/agents/ |
| Codex | ~/.codex/skills/ and ~/.agents/skills/ |
~/.codex/agents/ |
An install never replaces a personal file or an external link that occupies a managed name. It leaves the collision in place and reports it.
Check Status
Section titled “Check Status”hyprlayer ai statusShows the bundle version, where each harness is installed, and the config file path.
hyprlayer ai status --jsonAdds assetsVersion, pinnedVersion, binaryVersion, desiredVersion, and a platforms array for scripting.
Updates
Section titled “Updates”The bundle is pinned to a release version. Each command compares the installed bundle version with the wanted one, which is the binary’s own version unless you pinned another, and reinstalls only on a mismatch. When they match, startup does no network I/O.
To skip the daily release check, set disableUpdateCheck: true at the top level of your configuration file.
Pin or Reinstall
Section titled “Pin or Reinstall”hyprlayer ai versions # releases that carry a bundlehyprlayer ai reinstall # repair the current generationhyprlayer ai reinstall --force # download a fresh bundlehyprlayer ai reinstall --version 1.6.0 # pin to a version and install ithyprlayer ai reinstall --unpin # back to the binary's own bundleA pin survives binary upgrades, so you can hold a bundle back until a regression is fixed. A pin is refused if its bundle needs a newer CLI than the one running.