add an Agent Skill (agentskills.io spec) for using the mcpls CLI

[bug-ops]

Description

mcpls is designed to be driven by AI coding agents over MCP, but there's currently no packaged, spec-compliant "skill" that teaches an agent how to install, configure, and invoke the mcpls CLI itself (as opposed to the MCP tools it exposes once running). Add a skills/mcpls/ directory containing a SKILL.md that follows the Agent Skills specification, so agents that support the skill format (e.g. Claude Code, or any client implementing the spec) can discover and use it directly from this repo.

Proposed Skill Structure

Per the spec, the skill must live in a directory whose name matches the name frontmatter field:

skills/mcpls/
├── SKILL.md          # required: frontmatter + instructions
├── references/        # optional: deeper CLI/config reference
│   └── configuration.md   # can point to docs/user-guide/configuration.md content
└── assets/             # optional: example mcpls.toml snippets

SKILL.md frontmatter should follow the spec's constraints:

• name: mcpls — lowercase, matches directory name
• description — what the skill does and when to use it (e.g. "Install, configure, and run the mcpls CLI to bridge MCP and LSP for code intelligence. Use when the user wants to set up mcpls, add a language server, or troubleshoot the mcpls bridge.")
• license: MIT OR Apache-2.0 (matches Cargo.toml)
• compatibility — note it wraps a Rust binary (mcpls) and requires at least one LSP server installed
• metadata — repo/version info as needed

Body content should cover, at minimum:

• Installation (cargo install mcpls, pre-built binaries, cargo install --path crates/mcpls-cli)
• CLI flags exposed by crates/mcpls-cli/src/args.rs (--config/MCPLS_CONFIG, --trust-project-config/MCPLS_TRUST_PROJECT_CONFIG, --log-level/MCPLS_LOG, --log-json/MCPLS_LOG_JSON, --listen/MCPLS_LISTEN, --http-path)
• How to register mcpls with an MCP client (stdio and, if the transport-http feature is enabled, HTTP)
• Config file locations and the project-local mcpls.toml trust model (see docs/user-guide/configuration.md and the --trust-project-config security note in the README)
• Common troubleshooting pointers, linking to docs/user-guide/troubleshooting.md rather than duplicating it

Acceptance Criteria

• skills/mcpls/SKILL.md exists with spec-valid frontmatter (validated via skills-ref validate ./skills/mcpls if available, or manually against the naming/length constraints)
• name matches the parent directory name (mcpls)
• description is specific enough to trigger correctly (what + when), not generic
• SKILL.md body stays under ~500 lines / ~5000 tokens; anything longer moves to references/
• CLI flags and env vars documented match crates/mcpls-cli/src/args.rs exactly (avoid drift — consider a note or CI check to keep them in sync)
• No content duplicated verbatim from docs/user-guide/* where a reference/link would do
• README.md links to the new skill directory alongside the existing Documentation section

References

• Agent Skills specification: https://agentskills.io/specification
• Existing CLI args: crates/mcpls-cli/src/args.rs
• Existing user docs to draw from: docs/user-guide/configuration.md, docs/user-guide/installation.md, docs/user-guide/troubleshooting.md