local-skills-mcp

Provides a local MCP server that exposes filesystem-based “skills” (SKILL.md files) to any MCP-compatible LLM/agent client. It aggregates skills from multiple default directories and a custom SKILLS_DIR, supports lazy loading (names/descriptions first, full content on-demand), and hot reloads changes without restarting.

Evaluated Mar 30, 2026 (21d ago)
Repo ↗ Ai Ml mcp local-filesystem agent-tools prompt-skills lazy-loading hot-reload typescript
⚙ Agent Friendliness
55
/ 100
Can an agent use this?
🔒 Security
30
/ 100
Is it safe for agents?
⚡ Reliability
26
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
70
Documentation
55
Error Messages
0
Auth Simplicity
100
Rate Limits
0

🔒 Security

TLS Enforcement
20
Auth Strength
20
Scope Granularity
10
Dep. Hygiene
60
Secret Handling
50

Designed for local filesystem MCP usage with no authentication described. Hot-reload and multi-directory aggregation increase the risk of exposing unintended or maliciously modified skill files to an agent. TLS/auth controls for any network exposure are not described (likely local-only). No explicit secret redaction or logging guarantees are stated in the provided README.

⚡ Reliability

Uptime/SLA
0
Version Stability
45
Breaking Changes
40
Error Recovery
20
AF Security Reliability

Best When

You want local, private, prompt/skill reuse across multiple MCP clients with minimal context overhead and easy hot-reload.

Avoid When

You need strong transport/auth controls for exposing data to remote parties, or you run in environments where local filesystem access from an agent is too risky.

Use Cases

  • 讓 an agent reuse reusable expert prompt instructions stored locally
  • Context-efficient skill libraries for code review, writing, research, or operations
  • Portable MCP skill distribution across Claude Code/Desktop, Cline, Continue.dev, and custom MCP clients
  • Team/project-specific skill packs via local directories and overrides

Not For

  • Serving remote/shared skills to untrusted users over a network
  • Highly secure, multi-tenant environments without additional isolation/controls
  • Use cases requiring an authenticated, fine-grained API for external consumers

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
Yes
Webhooks
No

Authentication

OAuth: No Scopes: No

No authentication/authorization mechanisms are described for the MCP server; it appears designed for local use by a local MCP client.

Pricing

Free tier: No
Requires CC: No

Open-source MIT package distributed via npm; costs are primarily compute/inference from your LLM/agent, not this server.

Agent Metadata

Pagination
none
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • Skill selection depends heavily on the SKILL.md frontmatter description and trigger keywords; vague descriptions can lead to poor tool/skill choice.
  • The server aggregates multiple directories with override precedence; mismatched expectations about priority can cause the wrong skill version to load.
  • Since it reads local filesystem content, skills that include unsafe instructions may be surfaced to the model.

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for local-skills-mcp.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-03-30.

8642
Packages Evaluated
17761
Need Evaluation
586
Need Re-evaluation
Community Powered