Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
MCP server to list, load, explain, and validate Agent Skills.
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
The maker has not listed install details in the registry. See the maker’s README below.
Runs on your computer; the tool list appears in your AI app after you install it.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 7, 2026
The registry name was verified with the GitHub account craftbag, which owns the repo
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
This one runs on your computer; there is no internet endpoint
Runs on your computer and uses local settings instead
Runs on your computer; we can only read tool lists from internet endpoints
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
The registry does not list a required key, though some servers ask you to sign in on first use.
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
The GitHub repo last changed –, and the latest version (0.3.1) reached the registry 16 days ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Rust on your computer.
In the same category: Spec Driven Development, Remote Desktop Commander, AvaHR
Servers in the same category from other makers, sorted by stars.
Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Read hiring data and create inactive assessment and questionnaire drafts for human review.
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
Discover, list, load, and explain Agent Skills (SKILL.md) for CLI and MCP hosts.
craftbag walks project and home skill trees (.agents, plus optional vendor trees for Claude, Cursor, Grok, and Bline). It then:
list)load), or heading keys / one heading (load --outline, load --section KEY)why)validate)The same operations exist over MCP stdio (skills_list, skills_load, skills_why, skills_validate). Hosts can filter, rank, and load skills without taking a dependency on any one agent product.
macOS and Linux (Homebrew):
brew install craftbag/tap/craftbag
Windows (Scoop):
scoop bucket add craftbag https://github.com/craftbag/scoop-bucket
scoop install craftbag/craftbag
Both commands install craftbag and craftbag-mcp. The MCP host then runs craftbag-mcp (on macOS GUI apps, use the full path if PATH is empty: /opt/homebrew/bin/craftbag-mcp).
From a Rust toolchain (crates.io):
cargo install --locked craftbag-cli
cargo install --locked craftbag-mcp
Library dependency:
craftbag = "0.2"
From git (unreleased tip):
cargo install --locked --git https://github.com/craftbag/craftbag craftbag-cli
cargo install --locked --git https://github.com/craftbag/craftbag craftbag-mcp
MSRV is 1.85.
Default list walks cwd-to-git .agents/skills and $HOME/.agents/skills. Claude, Cursor, Grok, and Bline trees stay out of that walk until you pass --vendor. When .cursor/skills (or the matching directory for another vendor) is present and that vendor is off, the empty list names the flag: hint: .cursor/skills is present; pass --vendor cursor. This clone has no project .agents. When the .agents directories are absent, craftbag list exits 0 and says no skills were found. craftbag new NAME writes .agents/skills/NAME/SKILL.md with a matching name and a placeholder description. Point --path at the demo tree (same catalog as the demo GIF):
git clone https://github.com/craftbag/craftbag
cd craftbag craftbag list --no-implicit-roots --path demo/workspace/.agents/skills --catalog craftbag load review-pr --no-implicit-roots --path demo/workspace/.agents/skills craftbag load review-pr --no-implicit-roots --path demo/workspace/.agents/skills --outline craftbag load review-pr --no-implicit-roots --path demo/workspace/.agents/skills --section review-a-pull-request craftbag why review-pr --no-implicit-roots --path demo/workspace/.agents/skills --context review craftbag validate demo/workspace/.agents/skills/review-pr
--section review-a-pull-request passes the outline key. --outline prints that key before the heading Review a pull request. validate takes the package directory that contains SKILL.md, here demo/workspace/.agents/skills/review-pr. The collection folder demo/workspace/.agents/skills has no SKILL.md of its own, so validate on it exits 1 with directory is not a skill package (no SKILL.md). list --path on that folder still lists the packages inside it.
--path, --user-dir, and --ignore expand ~/ themselves, because MCP and a quoted command have no shell. ~//name stays under your home directory, the same as ~/name. A ~// remainder that is still a Windows drive path is refused, so it does not replace your home directory. A path that is not there is an unreadable skip and says the path does not exist. A symlink whose target is gone says dangling symlink. Explicit / and /.. are roots you named. A padded .. that collapses to /, such as /tmp/ .., is refused.
If craftbag is not on PATH yet, cargo build -p craftbag-cli --locked and use ./target/debug/craftbag in those commands.
In a project that already has skills under .agents/skills:
craftbag list --catalog
craftbag load NAME craftbag why NAME --context review craftbag validate ./path/to/my-skill
Claude, Cursor, Grok, or Bline trees are opt-in:
craftbag list --vendor claude --catalog
craftbag-mcp speaks JSON-RPC on stdio. Tools: skills_list, skills_load, skills_why, skills_validate. prompts/list and prompts/get expose skills whose user_invocable is true (the default) as MCP prompts. prompts/get returns the same text envelope as load. After brew install or scoop install, craftbag-mcp --help names the tools. A request with "id": null is answered with that id. A message that omits id is a notification and gets no response.
Claude Desktop (claude_desktop_config.json) and other hosts that take a stdio command:
{
"mcpServers": { "craftbag": { "command": "craftbag-mcp", "args": ["--vendor", "claude"] } } }
Launch --path, --vendor, --user-dir, and --no-implicit-roots are the walk when a tool call omits that field. The host cwd is still the implicit walk root unless you pass --no-implicit-roots. skills_load accepts outline and section (same as load --outline / --section KEY).
use craftbag::{discover, DiscoveryOptions};
fn main() -> std::io::Result<()> { let cwd = std::env::current_dir()?; let report = discover(&cwd, &DiscoveryOptions::default()); for skill in &report.skills { println!("{} {}", skill.name, skill.description); } Ok(()) }
implicit_roots is on by default (cwd-to-git .agents and $HOME/.agents). Set it to false and put collection roots in paths for leftover-only hosts. format_load_view can print an outline or one heading instead of the whole body.
See CONTRIBUTING.md. Security reports go to SECURITY.md. Roadmap and governance are in ROADMAP.md and GOVERNANCE.md.
Apache-2.0 or MIT. You may choose either. See LICENSE and LICENSE-APACHE.