plugins-claude/agentic-ide/skills/setup/SKILL.md
Install the agentic-ide tools: Serena MCP, Semgrep MCP, ast-grep CLI
npx skillsauth add st0nefish/claude-toolkit setupInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Status check first, then install instructions per tool. Run the bash block to see what's missing, then follow the section(s) for any ✗ entries.
echo "=== agentic-ide tool status ==="
echo
if command -v ast-grep &>/dev/null; then
echo "✓ ast-grep $(ast-grep --version 2>&1 | head -1)"
else
echo "✗ ast-grep not installed"
fi
if command -v serena &>/dev/null; then
echo "✓ serena $(serena --version 2>&1 | head -1)"
else
echo "✗ serena not installed"
fi
if command -v semgrep-mcp &>/dev/null; then
echo "✓ semgrep-mcp installed"
else
echo "✗ semgrep-mcp not installed"
fi
if command -v semgrep &>/dev/null; then
echo "✓ semgrep $(semgrep --version 2>&1 | head -1)"
else
echo "✗ semgrep not on PATH (semgrep-mcp scans will fail)"
fi
echo
echo "MCP servers must also be registered in ~/.claude.json — see sections below."
Both MCP servers install via uv (Astral's Python package manager):
curl -LsSf https://astral.sh/uv/install.sh | sh # universal
# or: brew install uv / pacman -S uv
Install:
uv tool install --from git+https://github.com/oraios/serena serena
Verify: which serena && serena --version. Upgrade later with uv tool upgrade serena.
Register in ~/.claude.json (user-level mcpServers, or per-project under .projects[<path>].mcpServers):
{
"mcpServers": {
"serena": {
"type": "stdio",
"command": "serena",
"args": [
"start-mcp-server",
"--context", "claude-code",
"--project-from-cwd",
"--disable-memories",
"--disable-onboarding"
],
"env": {}
}
}
}
--context claude-code tunes prompts and tool descriptions.--project-from-cwd auto-detects the project — no per-project config needed.--disable-memories --disable-onboarding skips Serena's opaque sidecar notes; project context belongs in source-controlled files.Reconnect Claude Code. mcp__serena__* tools should appear; mcp__serena__get_symbols_overview on a source file should return structured data. Logs at ~/.serena/logs/.
cargo install ast-grep --locked
# or: brew install ast-grep
Verify: ast-grep --version. No MCP registration — it's a plain CLI.
semgrep-mcp shells out to the semgrep scan engine at runtime, so the semgrep binary must be on PATH. uv tool install only links the primary tool's entry point, so install the engine's executable alongside it:
uv tool install semgrep-mcp --with-executables-from semgrep
# --with-executables-from also links the `semgrep` engine binary onto PATH
# equivalent: pipx install semgrep-mcp && pipx install semgrep
Verify: which semgrep semgrep-mcp && semgrep --version. Both binaries must resolve; if semgrep is missing, MCP scans will fail. Upgrade later with uv tool upgrade semgrep-mcp.
Register in ~/.claude.json:
{
"mcpServers": {
"semgrep": {
"type": "stdio",
"command": "semgrep-mcp",
"args": [],
"env": {}
}
}
}
Optional: add SEMGREP_APP_TOKEN to env to enable semgrep_findings (pulls from Semgrep AppSec Platform). Generate at https://semgrep.dev/orgs/-/settings/tokens. All local-scan tools work without it.
Reconnect Claude Code. mcp__semgrep__* tools should appear; mcp__semgrep__supported_languages should return a language list.
If MCP tools don't appear after registration:
which <command> — binary must be on PATH"type": "stdio" and the binary name as commandSemgrep is not installed or not in your PATH → the semgrep engine binary isn't on PATH; reinstall with uv tool install semgrep-mcp --with-executables-from semgrepdevelopment
Start work from your description — explore the codebase and plan
data-ai
Multi-phase, multi-agent feature workflow: spec → plan → refine → divide → execute → review. Invoke when the user escalates a session-start/session-issue flow to orchestration, or asks to run a non-trivial feature (multiple files, design ambiguity, cross-cutting concerns, correctness-critical paths) through the full multi-agent workflow. For small fixes, prefer session-start.
tools
Browse open issues, pick one, and start work on it
tools
Interact with GitHub and Gitea issue trackers and CI systems. List and show issues, file bugs, comment on issues or PRs, list and show pull requests, and fetch CI run logs — all from any repo context without leaving the session.