grounded-engineering
Reviewed
at commit 63dbe3a
Source-backed engineering practices and reviewed skill repositories, in a Claude Code pane.
0 upvotes
Log in to upvote or save- Made by
- Grounded Engineering
- Pinned commit
-
63dbe3a18fca337c9da789146ee5ab14f553eec2 - What we checked
-
Opens a Grounded pane listing engineering practices and reviewed skill repositories from a catalogue bundled with the plugin, ranked by what it finds in your project's files (CLAUDE.md, .claude settings, agents and skills, package manifests, CI). Its buttons start two skills that propose a change or explain an install and ask before writing or installing anything. The pane itself makes no network calls.
Review details
Static review of the source at the pinned commit; not run. register.tsx registers session.start (registers the /grounded-skills command and a hidden immediate grounded-link command, then reads the repository), command.run for grounded-engineering:grounded, grounded-skills and grounded-link, command.describe to hide grounded-link, adapt and explain from the menu, ui.message (presses from the desktop Client) and ui.render for the pane. Runtime calls: $.fs.read, $.fs.exists and $.fs.list on the bundled catalog.json and on project-relative paths only (.claude/skills/*/SKILL.md, .claude/agents/*.md, .claude/settings.json for hooks and sandbox.enabled, package.json, pyproject.toml, requirements.txt, go.mod, Cargo.toml, tsconfig.json, test directories, .github/workflows, .gitlab-ci.yml, .circleci/config.yml, .grounded-engineering/manifest.yaml), plugin state atoms for the pane's view settings and detected signals, $.ui.open, $.ui.toast and $.command.run. Client presses are validated against a fixed action table (adapt and explain only with ids that exist in the catalogue), and the link command only prints URLs the pane itself draws (practice evidence on the author's GitHub repository, listed skill repositories, two awesome lists) into the transcript. No fetch, process or model calls in the hook code; pressing Adapt or Explain starts a Claude turn on your account, which uses tokens. The adapt skill inspects the repository, shows a diff and writes only after approval; if the project already has the grounded-engineering CLI it runs npx --no-install grounded-engineering check, which cannot download anything. The explain skill has Claude read the listed repository on github.com at its pinned commit and quote the catalogue's install steps; only after an explicit yes does it ask you to type slash-command steps or run the shell steps (two are npx skills add commands, one unpinned skills@latest). No skill pre-approves tools, so Claude still asks permission for each command. Not included and not loaded by the code: fit-rules.yaml and hooks/test-support.ts.
- What it uses
-
- Reads your prompts: no
- Sees tool calls: no
- Blocks or changes tool calls: no
- Reads or writes files: yes
- Runs programs on your machine: no
- Calls the model (uses tokens): no
- Sends data over the network: no
- Stores data: yes
- Shows something on screen: yes
- Hooked events
Made this mod? Add a badge to your README
Paste this into your README. The badge says reviewed once the current version has been reviewed, and links readers to this page.
[](https://modsforclaude.com/mods/grounded-engineering/)