⏳ This skill is pending AI review.
Scores will appear once the review pipeline completes.
explain-error
Explain a stack trace or error message and suggest concrete fixes.
Choose how to use this skill
You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.
1. Native installer
This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.
Do not guess an installer command or replace an existing version without reviewing the diff.
2. Complete package recommended
Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.
No complete ProSkills package is published for this listing yet.3. Prompt-only
Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.
Need only the instruction file?
Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.
No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.
// RATINGS
Not yet listed on ClawHub or SkillsMP
// README
skillet
A tiny, zero-dependency package manager for MCP servers, Agent Skills, and Claude Code commands, hooks, subagents, and output styles.
Search a curated registry and install MCP servers, skills, slash commands, hooks, subagents, and output styles with one command - into Claude Code, Cursor, Codex, Gemini CLI, Copilot, and Cline.
skillet installs MCP servers (Model Context Protocol), Agent Skills (the open
SKILL.md standard), and Claude Code slash commands, hooks, subagents, and output styles into
Claude Code, Cursor, OpenAI Codex, Gemini CLI, Copilot, and Cline - from a curated, verified
registry of 370+ entries. One static Go binary, no Node or npm.
$ skillet search example
NAME DESCRIPTION TAGS
hello-skill Minimal example skill that ships with skillet - install i... example,starter,docs
$ skillet add hello-skill
Installing hello-skill from https://github.com/Brattlof/skillet ...
Installed hello-skill -> ~/.claude/skills/hello-skill
$ skillet add hello-skill --target agents
Installed hello-skill -> ~/.agents/skills/hello-skill # Cursor, Codex, Gemini, Copilot
$ skillet list
NAME KIND INSTALLED SOURCE STATUS
hello-skill skill a1b2c3d github.com/Brattlof/skillet tracking
Why use skillet?
Copy-pasting skill folders by hand, hunting through random repos, never knowing what's worth
installing. skillet fixes that: search a curated registry and install with one command.
- One static binary, no dependencies. The registry ships embedded - no backend, no account.
- Curated from official and widely-used sources, with every link and package verified. No dead-link slop. Contributed entries follow the one rule: you've actually used it.
- Skills use the open Agent Skills
SKILL.mdformat, so they work across tools. Install for Claude Code (default) or for the shared~/.agents/skillsdirectory with--target agents, which Cursor, Codex, Gemini CLI, and Copilot all read.
Supported tools
skillet installs for Claude Code by default (~/.claude/skills, plus its commands, hooks,
subagents, and output styles). Add --target agents and skills go to the shared
~/.agents/skills directory that Cursor, OpenAI Codex, Gemini CLI, and GitHub Copilot all read.
MCP servers install into each client's own config (Claude Code, Cursor, Windsurf, Gemini CLI,
and Cline). Commands, hooks, subagents, and output styles are Claude Code specific; skills and
MCP servers are the cross-tool part.
Install
# Go users
go install github.com/Brattlof/skillet/cmd/skillet@latest
# Or grab a binary from Releases
# https://github.com/Brattlof/skillet/releases
Building from source:
git clonethis repo, thengo build -o skillet ./cmd/skillet.
Usage
skillet add <name>[@ref] Install any registry entry (optionally pinned)
skillet add <owner/repo/skill>
Install a community skill from skills.sh
skillet install Restore everything from skillet.lock
skillet install --frozen Verify the install matches skillet.lock (no changes)
skillet lock Write skillet.lock from what is installed
skillet update [name] Update an installed item, or all of them
skillet doctor Check installed items for problems
skillet remove <name> Remove an installed item
skillet list List installed items
skillet search <query> Search the registry, then skills.sh (--kind, --tag)
skillet info <name> Show details of a registry entry or community skill
skillet trending Show trending skills on skills.sh (--limit N)
skillet hot Show skills rising on skills.sh in the last hour
skillet registry Show every registry entry
skillet publish How to publish your own skill
skillet completion <sh> Output a bash, zsh, or fish completion script
skillet self-update Update the skillet binary to the latest release
Add --target agents to any command to use ~/.agents/skills instead of
~/.claude/skills, or set SKILLET_TARGET=agents. skillet never overwrites a file or
folder it did not install; pass --replace to add or install to allow it.
Skills, commands, hooks, agents, output styles, and MCP servers
A registry entry has a kind: skill (the default), command, hook, agent,
output-style, or mcp. skillet installs each into the place the tool reads:
- a skill is a folder copied to
~/.claude/skills/<name>/(or~/.agents/skills/<name>/with--target agents, the shared location Cursor, Codex, Gemini CLI, and Copilot read), - a command is a single
.mdfile copied to~/.claude/commands/<name>.md, - a hook is a script copied to
~/.claude/hooks/, and skillet registers it in~/.claude/settings.jsonunder the event from the entry'shookblock (and un-registers it onremove), - an agent (subagent) is a single
.mdfile copied to~/.claude/agents/<name>.md, - an output style is a single
.mdfile copied to~/.claude/output-styles/<name>.md, - an MCP server is registered into a client's MCP config;
--targetpicks the client.
Commands, hooks, agents, and output styles are Claude Code specific. Skills and MCP servers work across tools:
skillet add context7 --target cursor # -> ~/.cursor/mcp.json
skillet add context7 --target windsurf # -> ~/.codeium/windsurf/mcp_config.json
skillet add context7 --target claude # -> ./.mcp.json (project scope)
skillet remove context7 --target cursor
Supported MCP clients: claude, cursor, windsurf, gemini, cline. skillet merges the server into the existing config and leaves every other entry untouched.
Shell completion
# bash (current shell)
source <(skillet completion bash)
# zsh
skillet completion zsh > "${fpath[1]}/_skillet"
# fish
skillet completion fish > ~/.config/fish/completions/skillet.fish
Completion suggests subcommands, the --target and --dir flags, and skill names pulled
from the local index, so it stays instant and works offline.
Choose where skills install, per-command or globally:
skillet add hello-skill --target agents # ~/.agents/skills (cross-tool)
export SKILLET_TARGET=agents # make it the default
skillet add hello-skill --dir ~/somewhere/else # an exact directory
Community skills from skills.sh
The curated registry is small on purpose. For everything else, skillet also searches skills.sh, Vercel's open directory of Agent Skills:
$ skillet search pdf
NAME KIND DESCRIPTION TAGS
pdf skill Fill, create, merge, split, and extract ... pdf,documents
From skills.sh (community, not reviewed by skillet):
SKILL INSTALLS
openai/skills/pdf 13.6k
github/awesome-copilot/pdftk-server 10.0k
$ skillet add openai/skills/pdf
audit: Gen Agent Trust Hub pass, Socket pass, Snyk warn (Risk: MEDIUM · 1 issue), ...
Installed open
// HOW IT'S BUILT
// FILES
2 files shipped with this skill. Inline text is the catalog copy. The upstream link is the source repository.
examples/explain-error/SKILL.md706 Bupstream ↗
--- name: explain-error description: Explain a stack trace or error message and suggest concrete fixes. --- # explain-error Use this when the user pastes an error, a stack trace, or failing build output. 1. Find the actual failure: the first real error, not the cascade of messages after it. 2. Explain in plain language what it means and the most likely cause. 3. Point at the specific file and line when the trace gives one. 4. Suggest one or two concrete fixes, with the exact change to make. 5. If the cause is ambiguous, say what to check or run next to narrow it down. Prefer the smallest fix that addresses the root cause over a workaround, and say so when you are guessing rather than certain.README.mdupstream ↗
Contents are not inlined in the catalog. Open the upstream file.
Plain-text copy, no login: /skill-bundles/brattlof-skillet-examples-explain-error.md