skillZs
★ LIVE SKILL TAGS ★
>>> LIVE SKILLS INDEX <<<
* OPEN SOURCE *
NO LOGIN, NO TRACKING
※ REAL INSTALL DATA ※
← back to all skills
sjunepark/agent-scripts85 installs

skills-cli

Skills CLI (`bunx skills`, skills.sh) for Codex, Claude Code, and Pi: discover, initialize, install, list, update, remove, restore lock files, sync `node_modules`, or troubleshoot project/global scope.

How do I install this agent skill?

npx skills add https://github.com/sjunepark/agent-scripts --skill skills-cli
view source ↗

Is this agent skill safe to install?

  • Gen Agent Trust Hubpass

    This skill provides a management interface for AI agent skills, allowing discovery, installation, and reconciliation of functional extensions across different agent platforms. It facilitates downloading code from remote repositories and modifying user-level configuration directories. Security features like SHA256 integrity checks, recovery manifests, and quarantine zones for replaced files are documented to mitigate risks associated with managing executable instructions.

  • Socketwarn

    1 alert: gptAnomaly

  • Snykwarn

    Risk: MEDIUM · 1 issue

What does this agent skill do?

Skills CLI

Workflow

  1. Inspect current state with bunx skills list for project scope and bunx skills list -g for global scope.
    • In this repository, use skill-registry.json as the authoritative classification, sjskills plan for committed project intent, and sjskills plan --global for exact fixed-baseline state.
  2. Use bunx skills find <query> or bunx skills add <source> --list to discover options.
    • When choosing a command, source, scope, or agent target—including initializing a skill—read references/cli.md for the CLI's accepted inputs and install-location facts.
  3. Install with explicit scope, skill selection, and agent targeting.
    • Before installing or reinstalling a published skill, or moving a shared install to agent-specific paths, read recipes/install-and-migrate.md and use the matching recipe.
    • In this repository, prefer sjskills over hand-written commands for both committed project intent and the fixed global baseline. Its private adapter materializes with explicit Codex and Claude Code targets; sjskills owns verified placement, state, quarantine, and restore.
    • Byte equality alone does not grant reconciler ownership. Managed updates and removals proceed only while current content matches trusted state.
  4. Verify ordinary installs with the matching scope command. For this repository's managed state, verify exact placement with sjskills plan for a project or sjskills plan --global for the fixed global baseline; shared-root discovery may report incidental agent visibility.
  5. Maintain or remove installs with the same scope and agent targeting used to create them.
    • Before listing filtered installs, removing skills, running generic updates, restoring a lock file, or syncing package-provided skills, read recipes/manage-installs.md.
    • For this repo's published skills, publish local changes before reinstalling from the remote source.

Guardrails

  • Default to symlink mode; use --copy when the user requests it or for a scoped global install of selected repo skills.
  • Use --skill '*' only when the user explicitly wants every skill from a source. In the current skills CLI, --all expands to --skill '*' --agent '*' -y, which can unintentionally recreate shared ~/.agents/skills installs.
  • Never synthesize a Pi target for this repository's fixed global baseline; Pi discovers the selected shared-root copy.
  • Publish changed repository skills before reconciliation. A local edit or unmerged branch is not present at a registry source pinned to main.
  • Do not run sjskills apply --global or global restore against a real home without a separately reviewed rollout plan and explicit authorization. Global apply also requires the reviewed JSON plan through --approved-plan and its approved digest through --approved-plan-sha256; those flags bind evidence but do not grant authorization.
  • Treat installed skills as executable instructions; avoid untrusted sources.
  • If managing dotfiles with chezmoi, avoid chezmoi add on live skills directories.

Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.

<a href="https://skillzs.dev/skills/sjunepark/agent-scripts/skills-cli">View skills-cli on skillZs</a>