agentleFS
rahmanef63rahmanef63/si-coder-agent/skills/sc-help/SKILL.md

sc-help

Skill13 starsChanged 2 months agoMITView on GitHub

Quick reference for SI-Coder route selection, frontend quality presets/profiles, secret-safe provider control, portable installation, and provider-specific skills. Use for 'sc help', 'what should I run', 'which frontend preset', 'which deploy route', or 'list si-coder commands'.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep durable instructions in English.

    line 16
  2. Distinctive:

    One user action at a time: Never dump a setup checklist when only one permission/account connection blocks progress.

    line 31
  3. Distinctive:

    Never ask the user to copy values between services: when a connector/server-side flow can do it safely.

    line 35
  4. Very specific:

    Every completion message states what works and offers exactly one [rekomendasi] next step.

    line 37
  5. Very specific:

    When a planner/tool returns userPlan, use it as the default user-facing response and keep raw routing/credential internals advanced.

    line 39
  6. Very specific:

    If a custom API-key provider has no creation URL, do not guess one.

    line 183
  7. Distinctive:

    Never put the credential value in chat, argv, logs, recommendations, or tool JSON.

    line 184
  8. Distinctive:

    After every meaningful completed milestone, emit exactly one next-step block:

    line 186
  9. Distinctive:

    Do not dump multiple recommendations.

    line 196

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

sc-help

Language

Keep durable instructions in English. Reply in the user's language unless they request another language.

Non-technical default UX — mandatory

SI-Coder is primarily for people who want a working web app, not an infrastructure lesson. Lead with the outcome, hide the plumbing.

A valid user request can be as simple as:

"Create a salon booking app and put it on my domain."

From that sentence, the agent should normally choose the stack, database/data service, hosting route, repository strategy, deployment method, domain records, frontend quality route, and verification approach itself.

Rules:

  1. Speak in goals: "publish the app", "improve the frontend", "connect the account", "connect the domain", "store the app data". Do not lead with environment variables, DNS records, deploy keys, containers, or internal skill routing.
  2. One user action at a time. Never dump a setup checklist when only one permission/account connection blocks progress.
  3. Do not ask users to choose technology unless they explicitly care. Choose sensible defaults and keep technology names in optional technical details.
  4. Do not ask a question that tools/repo state can answer. Inspect first.
  5. Credentials are framed as permissions, not secrets.
  6. Never ask the user to copy values between services when a connector/server-side flow can do it safely.
  7. Progress is product-oriented: Build → Frontend quality → Prepare data → Publish → Connect domain → Verify.
  8. Every completion message states what works and offers exactly one [rekomendasi] next step.
  9. Technical users can ask for --technical, JSON, or provider-specific skills.
  10. When a planner/tool returns userPlan, use it as the default user-facing response and keep raw routing/credential internals advanced.

Pick the entry point

GoalInternal skill identity
Describe anything you want SI-Coder to build/changesc
New/vague app ideasc-build
Combined frontend UI + UX + DX + AX worksc-fe
Visual interface/design-system/anti-slop auditsc-ui
Usability/accessibility/interaction auditsc-ux
Frontend developer-experience auditsc-dx
Agent Experience/tool/project ergonomics auditsc-ax
Find/install/create an Agent Skill by capabilitysc-skill
Existing app: publish from repo to productionsc-all
API/provider credential or account connectionsc-provider
Install in Claude Code/Codex/Hermes/OpenClawsc-install
Provider-specific operationmatching sc-* skill

Invocation is a host concern: slash-capable hosts can expose /sc-fe; ChatGPT Web uses automatic selection or the product's explicit skill/plugin selection UX rather than assuming custom slash registration.

Frontend presets and profiles

sc-fe preserves a coherent existing design by default. Use a named preset only when a direction is useful; presets encode principles, never proprietary assets or pixel clones.

/sc-fe --apple improve settings
/sc-fe --workbench audit the desktop shell
/sc-fe --profile baton-desktop refine the inspector
/sc-fe --save-profile baton-desktop
/sc-fe --apple --density compact --motion subtle
/sc-fe --audit --strict

Built-in shorthand presets: --apple, --workbench, --linear, --notion, --vercel, --material, --editorial, --terminal.

Key flags:

  • --preset <name> — explicit preset.
  • --profile <name> — load .sc/frontend/profiles/<name>.json.
  • --save-profile <name> — extract current design DNA into a reusable compact profile when writes are available.
  • --existing — explicitly preserve existing design DNA (default for a coherent existing product).
  • --fresh — allow a new direction.
  • --density compact|comfortable|spacious.
  • --motion none|subtle|expressive.
  • --platform desktop|mobile|responsive.
  • --audit — audit/report-first behavior.
  • --strict — applicable frontend verification failures block completion.

Explicit scope exclusions always win. A command such as "desktop shell only; do not touch mobile nav dock" must preserve that mobile surface even if shared styles/components are involved.

Deploy routing

# Hosted Claude Web / ChatGPT-style runtime
sc deploy plan --runtime hosted --composio

# Local runtime: VPS is the first branch
sc deploy plan --runtime local
  • hosted web/chat → full Composio for GitHub + Convex Cloud + Vercel + Hostinger; no VPS required.
  • local, VPS unknown → ask once whether the user has a VPS instead of guessing.
  • local + VPS → SC GitHub/Dokploy/self-hosted Convex.
  • local + no VPS → SC GitHub; Vercel/Convex/Hostinger prefer Composio, SC fallback.

Advanced deployment overrides: --runtime hosted|local and --target dokploy|hybrid|vercel|vps|managed. Frontend flags are routed to sc-fe, not interpreted as deployment flags.

Local CLI navigation

On a TTY, bare sc is a Finder-style alternate-screen TUI: navigation redraws one frame instead of appending terminal lines, visible SECTIONS tabs and a PATH breadcrumb stay at the top, and parent/current layers appear as columns. Use Tab/→ to enter a deeper branch, Enter to open or run, and ←/Esc to go back. Completing an action returns to the same TUI frame. Esc at Home does not quit; choose Quit or press Ctrl-D. If a credential or metadata input is open, Esc cancels only that input, saves nothing, and returns immediately to the previous Finder screen. On a normal 28-row terminal the lower INFO/PREVIEW/RESULT panel reserves 8 detail rows (6 on medium, 4 on compact terminals) so guidance uses the available vertical space.

Credential management is user-first. Always select a user before a provider/credential:

sc user
sc user show <user>
sc user duplicate <source> <target>
sc user connections <user> [provider]
sc user connection-add <user> <provider> "<label>" --source <sc|composio|native-mcp> --auth <method>
sc user credentials <user> [provider] [--connection alias]
sc user credential-set <user> <provider> [KEY] --connection <alias>
sc user credential-rm <user> <provider> [KEY] --connection <alias> --yes
sc user use <user>
sc user map <folder> <user>
sc user which

Finder hierarchy: Users → <user> → Providers → <provider> → Connections → <label> → Credentials → <KEY>. A GitHub direct connection (source=sc) uses classic-pat only and points to https://github.com/settings/tokens/new; GitHub through Composio remains a separate OAuth source. One user can own several isolated provider connections with unique labels, source/backends, auth methods, and scopes. Duplicate creates an independent user+connection tree. Never expose plaintext credential values.

For AI agents, prefer the matching sc.user.* MCP/machine tools from machine/functions.json. Use sc.user.connection.request/manage/list for account/auth selection and sc.user.credential.request for direct create/rotate handoff because machine JSON never accepts raw secret values.

Secret-safe commands

CommandPurpose
sc providers [--json]provider metadata + safe credential state
sc secret list/get ...state/source only; plaintext disabled
sc user credential-set <user> <provider> [KEY] --connection <alias>hidden local credential entry in one labeled connection
sc secret rm ... --yesremove managed credential
sc run --connection provider=alias -- <cmd>run child with explicit non-default connection, without printing secrets
sc doctorlive provider validation
sc audit --jsonmetadata-only lifecycle audit
sc update --check / sc updatesafe fast-forward self-update

Portable install

bash install.sh --agent claude
bash install.sh --agent codex --with-mcp
bash install.sh --agent hermes
bash install.sh --agent openclaw
bash install.sh --agent all

Claude plugin development/direct use:

claude --plugin-dir /path/to/si-coder-agent

Skills

Active default surfaces are catalog-driven and currently include /sc, /sc-build, /sc-all, /sc-fe, /sc-ui, /sc-ux, /sc-dx, /sc-ax, /sc-provider, /sc-install, /sc-help, /sc-git, /sc-dokploy, /sc-convex, /sc-convex-cloud, /sc-vercel, /sc-cf, /sc-onboarding, /sc-sync, and /sc-n8n. Dedicated Resend/Stripe/Clerk/Supabase automation is not installed/routed by default while unfinished. Their credential schemas may still be prepared or inspected through sc-provider; Cloudflare DNS is active.

After completing a task

Recommend one useful next step with its benefit + prerequisites, then ask whether the user wants it. Never request a raw secret in chat; use a secure external connection or connection-scoped hidden terminal handoff.

Mandatory credential + next-step response contract

Whenever a credential/API key is missing, never output only the variable name. Always make the handoff explicit:

Buat di      : <authoritative provider URL / secure connector auth link>
Petunjuk     : <minimum scope / exact menu when useful>
Connection  : <user/provider/label + scope>
Save with   : <sc user credential-set user provider KEY --connection alias, or provider connector>
Stored in   : <named SC connection 0600, or external connected account>
Lanjut       : <verification/resume action>

Rules:

  • Local SC runtime: resolve/create a labeled connection first, then use the provider endpoint and sc user credential-set <user> <provider> <KEY> --connection <alias>. Values live only in that connection's 0600 file; legacy profiles are migration fallback.
  • Hosted Claude Web/ChatGPT-style runtime: prefer the secure Composio connection URL returned by the connector; credentials stay in the connected account. Do not ask for the raw provider key unless the connector explicitly requires an API key bootstrap.
  • If a custom API-key provider has no creation URL, do not guess one. Require its provider metadata to be updated with --url https://... first.
  • Never put the credential value in chat, argv, logs, recommendations, or tool JSON.

After every meaningful completed milestone, emit exactly one next-step block:

[rekomendasi]
Next        : <one highest-value next step>
Why         : <one sentence>
Needs       : <prerequisites, or "nothing from you yet">
If you want : <what SI-Coder will do next / secure auth handoff>

Do not dump multiple recommendations. Do not recommend something already configured and healthy.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.