skrills

mcp
Guvenlik Denetimi
Gecti
Health Gecti
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 58 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose
This tool acts as a coordination and synchronization engine for AI coding assistants. It validates, analyzes, and syncs custom skills, commands, and configurations across multiple CLIs, and can be run as a local MCP server.

Security Assessment
The overall risk is rated as Low. The automated code scan checked 12 files and found no dangerous patterns, no hardcoded secrets, and no requests for excessive permissions. The tool does inherently parse local session history and access configuration files to perform its syncing duties, but it uses file hashing specifically to avoid overwriting user changes. Network access appears limited to fetching package data and connecting to configured MCP servers via stdio or HTTP transport.

Quality Assessment
Project health is excellent. It is licensed under the permissive MIT license and is highly active, with the most recent code push happening today. The repository has accumulated 58 GitHub stars, indicating a fair level of community trust and adoption. Additionally, the project demonstrates strong engineering hygiene by maintaining dedicated CI, test coverage, and security audit workflows.

Verdict
Safe to use.
SUMMARY

Coordinate skills between Codex, Copilot, and Claude Code. Validates, analyzes, and syncs skills, subagents, commands, and configuration between multiple CLIs.

README.md

Skrills

Skrills Icon

Crates.io
Downloads
Docs
CI
Coverage
Audit
License: MIT

Skills support engine for Claude Code, Codex CLI, GitHub Copilot CLI,
and Cursor.

Installation |
User Guide |
CLI Reference |
MCP Tutorial |
FAQ |
Changelog

What's new in 0.7.3 -- MCP tool filtering
(allowed_tools/disabled_tools) synced across all adapters,
MCP servers panel in TUI and browser dashboards,
/api/mcp-servers REST endpoint, configurable Unpaywall email,
OffsetDateTime for discussion timestamps, and SQL upsert fix.
See changelog.

Demo

Skrills Demo

TUI dashboard -- navigate skills, activity, and metrics with
keyboard shortcuts:

TUI Dashboard

See the quickstart tutorial for a
walkthrough, or the MCP tutorial for server
setup.

Features

  • Cross-CLI validation -- validates skills against Claude Code
    (permissive), Codex CLI (strict), Copilot CLI (strict), and Cursor
    rules. Auto-derives missing YAML frontmatter from file paths and
    content.
  • Multi-directional sync -- syncs skills, commands, agents, MCP
    servers, rules, and preferences across all four environments. Uses
    file hashing to respect manual edits so user changes are not
    overwritten.
  • Token analytics -- measures token usage per skill and suggests
    reductions to fit context windows.
  • Dependency resolution -- resolves skill dependencies with cycle
    detection and semantic versioning constraints.
  • MCP server -- 27 tools for validation, sync, intelligence, and
    project-aware skill generation over stdio or HTTP transport.
  • Session mining -- parses Claude Code and Codex CLI session
    history to improve recommendations based on actual usage.
  • Visualization -- TUI and browser dashboard showing discovered
    skills, validation status, usage metrics, and MCP server
    configurations with tool filter indicators. The browser dashboard
    supports light and dark modes. The standalone
    skrills-portal.html can be opened directly
    in a browser or uploaded to integration platforms without a running
    server.
  • Discovery deduplication -- frontmatter identity matching
    consolidates the same skill installed in multiple locations.

Installation

macOS / Linux:

curl -LsSf https://raw.githubusercontent.com/athola/skrills/HEAD/scripts/install.sh | sh

Windows PowerShell:

powershell -ExecutionPolicy Bypass -NoLogo -NoProfile -Command ^
"Remove-Item alias:curl -ErrorAction SilentlyContinue; iwr https://raw.githubusercontent.com/athola/skrills/HEAD/scripts/install.ps1 -UseBasicParsing | iex"

crates.io: cargo install skrills

See installation guide for HTTP transport
setup, systemd services, and advanced options.

Quickstart

# Validate skills for Codex/Copilot/Cursor compatibility
skrills validate --target both --autofix

# Analyze token usage
skrills analyze --min-tokens 1000 --suggestions

# Sync from Claude to all other CLIs
skrills sync-all

# Sync between specific environments
skrills sync --from cursor --to claude

# Start MCP server and open the browser dashboard
skrills serve --http 127.0.0.1:3000 --open

# Interactive TUI dashboard
skrills tui

# Launch an agent with automatic backend routing (Claude -> Codex fallback)
skrills multi-cli-agent my-agent

See CLI reference for all commands including
skill lifecycle management.

Supported Environments

Skrills syncs seven asset types across four CLI environments.
Each cell reflects what the adapter reads and writes today:

Asset Claude Code Codex CLI Copilot CLI Cursor
Skills Y Y Y Y
Commands Y Y -- Y
Agents Y -- Y Y
MCP Servers Y Y Y Y
Hooks Y -- -- Y
Instructions / Rules Y -- Y Y
Preferences Y Y Y --

Cursor rules (.mdc files) are mapped bidirectionally via mode
derivation (alwaysApply, glob-scoped, agent-requested).
See ADR 0006 for the
mapping strategy and sync guide for
workflows.

Skill Management

# Deprecate, rollback, import, or score skills
skrills skill-deprecate old-skill --replacement "new-skill"
skrills skill-rollback my-skill --version abc123
skrills skill-import https://example.com/skill.md
skrills skill-score

See CLI reference for the full set of
skill-* subcommands (skill-catalog, skill-profile,
skill-usage-report).

Architecture

Crate Purpose
cli Binary entry point (delegates to server crate)
server MCP server, HTTP transport, security middleware
validate Validation logic for Claude/Codex/Copilot/Cursor compatibility
analyze Token counting, dependency analysis, optimization
intelligence Recommendations, project analysis, skill generation
sync Multi-directional sync with adapters for each CLI (Claude, Codex, Copilot, Cursor)
dashboard TUI and browser-based skill visualization
discovery Skill discovery and ranking
state Environment config, manifest settings, runtime overrides
metrics SQLite-based telemetry for invocations, validations, sync
subagents Shared subagent runtime and backends
tome Research API orchestration, caching, PDF serving
test-utils Shared test infrastructure (fixtures, RAII guards, temp dirs)

See architecture docs for the crate
dependency graph and runtime flow.

Configuration

Create ~/.skrills/config.toml for persistent settings:

[serve]
auth_token = "your-secret-token"
tls_auto = true
cors_origins = "https://app.example.com"

Precedence: CLI flags > environment variables > config file.

See security docs for TLS setup (skrills cert
subcommand) and FAQ for environment variables.

Documentation

Resource Description
User Guide Primary documentation (mdBook)
CLI Reference All commands with examples
MCP Tutorial Server setup and tool reference
Sync Guide Cross-CLI sync workflows (Claude, Codex, Copilot, Cursor)
Token Optimization Context window management
FAQ Common questions
Security Auth, TLS, threat model
Changelog Release history

Limitations

  • No runtime skill injection: Skrills validates and syncs
    files; it does not inject skills into prompts at runtime.
  • Copilot command sync: Copilot CLI does not support slash
    commands, so command sync is skipped.
  • Cursor preferences: Cursor preferences are not yet mapped;
    preference sync is skipped for Cursor targets.
  • Empirical mining: Session history parsing works best with
    recent Claude Code / Codex CLI versions.
  • LLM generation: Requires ANTHROPIC_API_KEY or
    OPENAI_API_KEY for skill creation.

Plugin Marketplaces

Skrills validates, analyzes, and syncs skills from these exemplar
plugin marketplaces:

  • superpowers -- Opinionated
    skill pack for Claude Code covering TDD, code review, planning,
    debugging, and development workflows.
  • claude-night-market
    -- Plugin marketplace for Claude Code with skills, agents, hooks,
    and commands across multiple domains.

Development

make lint test --quiet

Requires Rust 1.75+. See development guide
for test coverage, CI, and contribution workflow.

Contributing

License

MIT

Star History Chart

Yorumlar (0)

Sonuc bulunamadi