Git-Fg

claude-cli-wrapper

Hub for the claude-cli-wrapper MCP server — execute, session, context, review, agent, and config tools. Routes to spoke skills.

agent-mgmt

Spawn, list, inspect, and stop sub-agents through the wrapper. Sub-agent config travels as a JSON-serialized string. Flat parameter surface.

Data & Documents Listed

config

Read, write, and validate the wrapper's local config file. Config blob is a JSON-serialized string. Source of truth: config-file.md.

context

Attach, replace, and inspect context payloads for Claude Code sessions. Complex structures travel as JSON-serialized strings.

execute

Execute Claude Code CLI invocations via the claude-cli-wrapper MCP server. Run prompts with mode, effort, output_format enums.

review

Run structured code review passes via the wrapper — flat params, mode/effort/output_format enums, return pass/fail findings as a string.

session

Manage Claude Code CLI sessions via the claude-cli-wrapper MCP server — create, list, resume, end. Session IDs are UUID v4.

diagnose

Investigate and find the root cause of complex bugs, recurring failures, or unexpected behavior. Use when user says 'why is this happening', 'find the bug', or 'analyze the failure'.

Plan or run a project end-to-end — PLAN mode creates executable PLAN.md roadmaps with checkpoints; EXECUTE mode runs a plan with worker + critic subagents. Use when the user wants to add a new feature, start a new project, break down work into phases, run a plan, or build something non-trivial.

project-maintenance

Archive completed plans, deduplicate stale memory, and clean auto-memory at the end of a project. Use when the user wants to wrap up a plan, clean up files, or run a memory cleanup pass.

refine

Review a PR, simplify complex code, polish prose, or capture a project learning. Use when the user wants to improve the quality of an artifact (code, doc, PR) or consolidate insights into memory. Do NOT use for bug diagnosis or for creating new features.

rules-orchestration

Manage CLAUDE.md and rules — extract learnings, add rules, restructure, and review. Use when updating project conventions.

security

Scan for security vulnerabilities, exposed secrets, and broken authentication patterns before production deployment. Use when user says 'security audit', 'dependency scan', or 'find secrets'.

Data & Documents Listed

session-analytics

Analyze session transcripts to extract data, diagnose anti-patterns, and generate GitHub issues from findings.

kaizen

Apply four design-time guardrails to every code decision: incremental improvement, error-proofing, standardization, and avoid over-engineering. Use when the user is about to write code, refactor, or handle errors and wants to avoid over-engineering or apply YAGNI.

ddd

Restructure tangled files, untangle mixed layers, fix business logic in the wrong place, audit an API contract, or resolve deeply nested code. Use when architectural decisions have grown irreversible, layers are backward, or responsibilities are scattered.

ideation

Explore a vague idea or generate creative alternatives before committing to an approach. Use when the user wants to 'think through', 'explore possibilities', or 'generate ideas' for an unformed problem.

plan-do-check-act

Execute evidence-based experimentation cycles to test hypotheses and measure the results of a change.

API & Backend Listed

skill-authoring

Create, optimize, and test Claude Code skills, including metadata refinement and trigger benchmarking.

subagent-orchestration

Design multi-agent architectures, author agent definitions, or orchestrate parallel execution. Use when delegating tasks to subagents.

task-lifecycle

Track and build features from initial idea to implementation. Use when the user wants to add a new feature, add a task, refine an idea, build something, or turn a rough description into a detailed technical spec.

test-orchestration

Plan test strategies and execute test-driven development cycles, including coverage addition and test repair.

web-search

Find, verify, and evaluate information on the open web. Use when the user wants to look something up, find the latest on a topic, research a question, fact-check a claim, evaluate whether a statement is true, or asks 'is this true?', 'where did this number come from?', or 'what do experts say?'. Covers query reformulation, source hierarchy, cross-reference verification, and the discipline of stopping when an answer is unfindable. Triggers on 'find X', 'look up Y', 'what is the latest Z', 'is this claim true', 'verify that', 'search for', 'research', 'current data on', 'best source for', 'I want a second opinion on X', 'is this real'.

fpf

Analyze a problem from first principles, generate and evaluate competing hypotheses, and track decision rationale. Use when the user wants to reason from scratch, compare solutions, or make a decision between alternatives.

git

Handle version control tasks: commit changes, review pull requests, and manage issues or branches. Use when user says 'commit', 'PR', 'issue', or 'branch'.

sadd

Solve a complex problem by generating multiple candidate solutions and picking the best one with structured evaluation. Use when the user wants to compare options, pick the best approach, or do a deep dive into a hard problem where a single attempt might fail.

claude-cli

Drive the Claude Code CLI as six MCP tools from another agent — claude_execute (run a prompt), claude_session (lifecycle — resume/continue/fork/list/info/close), claude_context (workspace — add dirs, worktrees), claude_review (code review), claude_agent (background agent management), claude_config (runtime tuning — model/effort/permissions/settings). Use when an agent needs to programmatically spawn or orchestrate Claude Code instances, manage their sessions, or run background sub-agents — not for direct user-driven Claude Code use.

git-worktree-manager

Manage git worktrees: create, list, remove, prune, and audit worktree state. Use when setting up parallel development environments, isolating branches for review, or managing multi-branch workflows.

mcp-server-design

Design an MCP server's tool surface — equilibrated recursivity (flat schema, deep data), 1 tool vs N tool decomposition, output contract, error-code discipline, capability negotiation, security checklist, tool naming, and the Claude-Optimal validation. Use when the user says "design an MCP server", "decompose MCP tools", "MCP schema design", "MCP tool naming", "MCP error codes", "MCP security checklist".

mcp-server-implement

Build an MCP server in Rust — tool attributes, server lifecycle (initialize → capabilities → shutdown), transport choice (stdio vs Streamable HTTP), stderr-only logging, error mapping, and testing with the MCP Inspector in CLI mode (no browser). Use when the user says "implement an MCP server in Rust", "build an MCP tool", "Rust MCP server", "transport choice stdio vs HTTP", "MCP server lifecycle".

rust-quality

Set up the Rust quality pipeline — CI with clippy, faster test runner (nextest), coverage reporting, dependency auditing, and the supply-chain ladder. Use when the user says "set up CI for Rust", "configure clippy", "speed up tests", "add coverage", "audit dependencies", "set up supply-chain", "add benchmarks", "lint as error in CI".

rust-release

Version, changelog, publish, and maintain a Rust crate — Cargo semver (including the contested MSRV policy), changelog tooling (git-cliff, release-please, hand-curated), the cargo publish playbook, the supply-chain maintenance (cargo-vet, Dependabot, RUSTSEC), and feature deprecation. Use when the user says "publish my Rust crate", "version my crate", "set up release-plz", "deprecate a feature", "bump MSRV", "set up cargo-vet".

Data & Documents Listed

rust-scaffold

Scaffold a new Rust project — lib/bin/lib+bin decision tree, the recommended Cargo.toml template (edition 2024, MSRV, metadata, profile.release), MSRV policy, the 5-rule feature flag playbook, the lib+bin code layout pattern, rustdoc conventions, examples & tests directory, and edition-2021-to-2024 migration. Use when the user says "new Rust project", "scaffold a crate", "Cargo.toml template", "lib + bin", "feature flag design", "edition 2024 migration".

rust-workspace

Decide on and structure a Rust workspace — single-crate vs workspace vs multi-workspace decision tree, virtual workspace template, the workspace inheritance playbook (1.64+), Cargo.lock policy, MSRV coordination across members, cross-crate patterns (internal features, path deps, shared dev-deps, feature unification), and workspace publishing. Use when the user says "set up a Cargo workspace", "split into a workspace", "share deps across crates", "workspace inheritance", "coordinate MSRV across members", "publish a workspace".

wiki

Search, query, ingest, or verify a personal wiki, knowledge base, or notes system. Use when user mentions 'wiki', 'KB', 'knowledge base', 'look up in my notes', 'find in my wiki', 'search my docs', 'lint wiki', 'check wiki consistency', 'add to wiki', 'ingest into wiki', 'populate wiki', or 'build wiki'.

git-preflight-checker

Run pre-flight validation checks (lint, type-check, tests, build sanity) before commits or PRs. Use before any commit or merge to gate quality. Exits fast with clear pass/fail so the main agent can proceed or abort.