Skip to content

dsh-developer v0.1.0

Choose a tag to compare

@builtin-pb builtin-pb released this 11 Sep 05:03
· 44 commits to main since this release

The first release of dsh-developer: one native workflow for building DSH plugins and working on the harness itself.

Included

  • Project and toolchain inspection, package-script execution, and navigation into the exact installed DSH declarations or an upstream checkout.
  • Packed-plugin installation and native tool verification in disposable profiles.
  • Web development with owned server lifetime, native authentication, and an optional isolated browser route configured through ui-setup.
  • Session diagnostics, static package checks, profile preflight and upgrade comparisons on the documented audit lanes.
  • Optional isolated Build/Apply with reviewed provider admission, exclusive file publication and preserved recovery state when cleanup or rollback cannot be verified.
  • One shared development skill for DSH and Codex, with native tool and Web examples.

Install

dsh plugin --profile web add 'github:builtin-pb/dsh-developer#v0.1.0' --ignore-scripts

Restart DSH Web after installation. Alternatively, download dsh-developer-0.1.0.tgz below and pass its local path to the same command. SHA256SUMS records the archive digest.

Requires Node.js ^22.18.0 || >=24.11.0. See the installation guide for setup.

Tested scope

All 15 cross-platform CI jobs passed. Release preparation passed 603 local tests, four real Mac VM checks, and both complete native DSH development suites. The release archive passed native verification and release/preview lifecycle checks.

Native development CI covers Windows, macOS and Linux with DSH 0.1.1-rc.2 and 0.1.5-rc.2. Audits and isolated Build have separate reviewed-runtime and provider requirements; current-runtime development support does not imply isolated execution support. The platform guide describes those boundaries.

Static analysis can report incomplete coverage for valid implementations. Such findings remain explicit; installation and passing tests do not establish correctness of arbitrary agent-generated code. Rendered browser checks were observed on macOS; rendered Windows browser automation remains outside that evidence.