Skillget

cskill-agents

mcp
Security Audit
Pass
Health Pass
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 11 GitHub stars
Code Pass
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested
Purpose
This project provides a curated collection of agent "skills" (configuration files and patterns) designed to enhance coding CLIs and multi-agent runtimes like Codex and Claude Code. It offers reusable instructions for building tool runners and context engines without copying or redistributing original source code.

Security Assessment
Overall Risk: Low. The repository is essentially a library of instructional text files and helper installation scripts rather than functional, compiled software. A light code scan of 12 files found no dangerous patterns, hardcoded secrets, or requests for dangerous permissions. The PowerShell and Bash installation scripts simply copy skill files to the appropriate agent directories. Because it is an installable tool, users should still quickly review the helper scripts before executing them on their local machine, but no active threats or data-access risks were identified.

Quality Assessment
The project is in active development, with its last push occurring today. It uses the permissive and standard MIT license, making it highly accessible for both personal and commercial use. While still an emerging tool with a modest 11 GitHub stars, the repository is very well-documented and cleanly organized. The author clearly defines the project's boundaries and explicitly states that no proprietary assets or third-party code are redistributed, which shows careful attention to legal and operational quality.

Verdict
Safe to use.
SUMMARY

Agent skills for coding CLIs, multi-agent runtimes, context engines, MCP extensions, and terminal tooling. Instead of using claude code's source code, give your agent skills to create your own!

README.md

cskill-agents

Curated, public-safe agent skills for building and improving coding CLIs, terminal tools, context engines, remote bridges, and multi-agent runtimes. (inspired from Claude Code's source)

This repo contains original, method-level skill files distilled from reviewed implementations. It does not ship upstream source code, copied comments, tests, prompts, logos, or proprietary assets.

Quick Start

Install one skill with the helper scripts. These installers validate the slug and replace any existing installed copy, so they are safe for upgrades and reinstalls.

Codex-compatible

.\scripts\install-skill.ps1 -Agent codex -Slug concurrency-safe-tool-batching

./scripts/install-skill.sh codex concurrency-safe-tool-batching

Claude Code-compatible

.\scripts\install-skill.ps1 -Agent claude-code -Slug concurrency-safe-tool-batching

./scripts/install-skill.sh claude-code concurrency-safe-tool-batching

Install a curated bundle with the helper script:

.\scripts\install-bundle.ps1 -Agent codex -Bundle starter
.\scripts\install-bundle.ps1 -Agent claude-code -Bundle starter

./scripts/install-bundle.sh codex starter
./scripts/install-bundle.sh claude-code starter

What You Can Build

  • Coding CLI command surfaces with dynamic skill loading, cache-safe registries, and curated slash-command discovery
  • Tool runners with concurrency-safe batching, ordered streaming, output persistence, and deterministic budgeting
  • Long-running agent sessions with compaction, memory prefetch, retry control, and stable session state
  • Multi-agent systems with isolated child contexts, scoped MCP clients, transcript sidechains, and lifecycle cleanup
  • Remote-safe bridges and terminal interfaces with capability filtering, privacy-aware init payloads, and motion-aware terminal UI

What This Repo Is

  • A curated distribution repo built from a larger authoring corpus.
  • A set of reusable skill files that tell agents how to reproduce strong patterns without copying implementation code.
  • A practical install target for Codex-compatible and Claude Code-compatible skill directories.
  • A bundle-based catalog of high-value agent-runtime patterns.

What This Repo Is Not

  • Not a source mirror.
  • Not a vendor-affiliated SDK, extension pack, or official template set.
  • Not a redistribution of third-party code, copied comments, tests, prompts, or internal assets.
  • Not the full authoring corpus; this repo only ships a curated public-safe subset.

Included Skills

  • 89 curated skills
  • 36 starter skills
  • 9 curated bundles

Start with:

Supported Agents

Compatibility here means the repo uses a skill-folder layout that these agents can consume. It does not imply endorsement, certification, or affiliation.

Curated Bundles

  • starter: best first install for most coding agents
  • command-surfaces: command registries, skill loaders, and capability surfaces
  • tool-orchestration: batching, streaming, persistence, and output shaping
  • context-management: compaction, memory, and session control
  • multi-agent: child-agent isolation, cleanup, and fork control
  • extensions-mcp: extension lifecycle and parity patterns
  • safety-worktrees: shell safety, path guards, and worktree handling
  • remote-bridge: remote-safe command and bridge transport patterns
  • terminal-ui: terminal presentation and feedback patterns

Repo Layout

cskill-agents/
  README.md
  LICENSE
  NOTICE
  LEGAL.md
  PROVENANCE.md
  CONTRIBUTING.md
  catalog/
  bundles/
  skills/
  agents/
  scripts/
  • skills/: canonical public universal skills
  • agents/: direct-install compatible skill folders
  • bundles/: curated install groups
  • catalog/: machine-readable and human-readable indexes
  • scripts/: install helpers and safety checks

Skill Format

Each public skill contains:

  • SKILL.md: the reusable method instructions
  • skill.yaml: sanitized public metadata for catalogs and agents

Public metadata is intentionally coarse. This repo does not publish raw source repo paths or internal mirror paths.

Public Provenance Model

Public artifacts in this repo use coarse provenance only:

  • pattern family
  • category
  • review status
  • bundle membership

Private maintainer evidence can exist outside this repo, but public files should remain abstraction-first and path-safe.

Legal And Non-Affiliation

Read:

Short version:

  • this repo licenses only its original authored materials
  • third-party product names are used only for nominative compatibility references
  • this repo is not affiliated with, endorsed by, sponsored by, or certified by Anthropic or OpenAI

Contributing

Before contributing, read CONTRIBUTING.md.

Public contributions must stay method-level. Do not add:

  • source code snippets
  • copied comments or tests
  • raw internal file paths
  • repo mirrors
  • proprietary assets

Removal Requests

If you believe a public artifact in this repo is too close to upstream protected material, open an issue or contact the maintainer. See LEGAL.md for the review and correction process.

Reviews (0)

No results found