Install
$ agentstack add skill-reopt-ai-reopt-skills-opt-shell-install ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
opt-shell Install
> This is NOT the opt-shell you know. @reopt-ai/opt-shell (formerly @reopt-ai/opt-harness) is the runtime product-frame layer between opt-ui and product screens. It ships no dist/docs/ — read node_modules/@reopt-ai/opt-shell/shell-llms.txt (agent guide) and README.md before writing code.
When to apply
Consumer project depends on @reopt-ai/opt-shell. Triggers: "install", "init", "setup", "upgrade", "update" — with shell / opt-shell (or the legacy harness / opt-harness). @reopt-ai/opt-harness was renamed to @reopt-ai/opt-shell; if a project still references opt-harness, migrate the dependency first.
What opt-shell provides
| Area | Pattern | |---|---| | Workspace recipes | DashboardWorkspace, ListWorkspace, DetailWorkspace, EditorWorkspace, landing — pick via the decision tree in shell-llms.txt | | Non-recipe surface | ShellFullscreenToolSurface (full-viewport tools — code editor, canvas) | | Policy | density, contentWidth (full / wide / normal / narrow), navigationMode, motion | | Adapters | data-engine wrappers supplying loading / empty / error chrome | | State UX | shared state boundaries; every recipe requires header + content slots | | Authoring audit | @reopt-ai/opt-cli/audit + opt harness commands; published opt-shell 1.0.0 has no ./audit export |
Invocation
/opt-shell-install # Auto-branch (missing → init, installed → upgrade)
/opt-shell-install --upgrade # Explicit upgrade
/opt-shell-install --check # Analyze only
Step 1 — Pin agent rules into AGENTS.md / CLAUDE.md
Source: the module's own agent-rules file once it ships one (@reopt-ai/opt-shell ships shell-llms.txt, an agent guide, but not a marker-block file as of 1.0.0). Fallback: agent-rules.md bundled with this skill. Wrap content between:
…content from source…
Idempotent: replace only between markers.
Step 2 — Consumer-side setup (this skill owns; docs cannot)
- Public npm registry — no token or scoped
.npmrcentry is required. Inspect the project.npmrcandnpm config get @reopt-ai:registry; if the scope still resolves to GitHub Packages, remove only the legacy project entry@reopt-ai:registry=https://npm.pkg.github.com. Preserve unrelated registry/auth settings, and ask before changing user/global npm config.
- Peers — opt-shell's
peerDependencies(install / run their skills first if missing):
- Required:
@reopt-ai/opt-palette(theme engine),react/react-dom19+ - Optional (only if you use that adapter):
@reopt-ai/opt-datagrid(/opt-datagrid-install),@reopt-ai/opt-editor(/opt-editor-install),@reopt-ai/opt-calendar
opt-ui tokens come in transitively through opt-palette — run /opt-ui-install if the theme layer is missing.
- App wiring — properties of the consumer app:
- A workspace recipe at the screen root (
header+contentslots are mandatory). - A shell manifest / policy config (density, contentWidth, navigationMode, motion).
- Adapters wired to your data engines wherever loading / empty / error chrome is needed.
Step 3 — Route to module docs
opt-shell ships no dist/docs/. Route to shell-llms.txt (agent guide) and README.md.
| Task signal | Read | |---|---| | Recipe decision tree, slot rules, policy | shell-llms.txt | | Component / prop API, exports (., ./core, ./meta) | README.md | | Authoring audit / scoring | @reopt-ai/opt-cli/audit; npx @reopt-ai/opt-cli harness check|test|doctor | | Breaking changes per version | CHANGELOG.md |
Pipeline (auto-branch)
| # | Step | Init | Upgrade | |---|---|---|---| | 1 | Detect current state (incl. legacy opt-harness dep) | ✓ | ✓ | | 2 | Public-registry preflight + legacy override cleanup | ✓ | ✓ | | 3 | Install / update package | ✓ | ✓ | | 4 | Peer check (opt-palette required; opt-datagrid / opt-editor / opt-calendar optional) | ✓ | ✓ | | 5 | Shell manifest / policy config | ✓ | – | | 6 | First workspace recipe | ✓ | – | | 7 | Breaking-change edits | – | ✓ | | 8 | Verify + summary | ✓ | ✓ |
Safety
- Never upgrade without an impact scan.
- Confirm before overwriting existing files.
- Apply breaking-change edits in logical groups, never bulk.
- Do not finish until
tsc --noEmitpasses. - Never commit — do not commit or push without an explicit request.
Verify
npx tsc --noEmitpasses.- App boots into the chosen workspace recipe (
header+contentslots filled); policy (density / contentWidth / navigationMode) applies. - (If an adapter is used) the data-grid or editor adapter renders the peer's content with shell chrome (loading / empty / error states).
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: reopt-ai
- Source: reopt-ai/reopt-skills
- License: MIT
- Homepage: https://skills.sh/reopt-ai/reopt-skills
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.