claude-forge: Skill for Claude Code

.claude/skills/forge-mcp-clients/SKILL.md

forge-mcp-clients is a skill for Claude Code from ForgeyClap/claude-forge. It costs 46 tokens per session (2,176 once invoked), scanned A, original, MIT.

A set of rules for using MCP servers, which provide tools that an AI agent can call, while keeping access optional and limited to approved permissions.

In plain words
What is it for?
Use it when deciding whether an MCP server may be used, checking opt-in and permission grants, or routing write operations through the project's existing approval gate.
Why use it?
It prevents an agent from silently installing or activating external tools and reduces the risk of granting more access than a task requires.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: mentions Codex.

This is ForgeyClap/claude-forge's own configuration. It tells Claude Code how to work on claude-forge itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything claude-forge configures →

Reuse

Borrowing it

Nothing to install: this file belongs to ForgeyClap/claude-forge. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/ForgeyClap/claude-forge/main/.claude/skills/forge-mcp-clients/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/ForgeyClap/claude-forge

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for forge-mcp-clients

README.md
[![agentmods](https://agentmods.dev/badge/skills/forgeyclap/claude-forge/forge-mcp-clients/github.svg)](https://agentmods.dev/skills/forgeyclap/claude-forge/forge-mcp-clients)
Your own site
<a href="https://agentmods.dev/skills/forgeyclap/claude-forge/forge-mcp-clients"><img src="https://agentmods.dev/badge/skills/forgeyclap/claude-forge/forge-mcp-clients/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for forge-mcp-clients

Your own site · 80×15
<a href="https://agentmods.dev/skills/forgeyclap/claude-forge/forge-mcp-clients"><img src="https://agentmods.dev/badge/skills/forgeyclap/claude-forge/forge-mcp-clients.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 46 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,176 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00046 $0.02176
Opus 5 $0.00023 $0.01088
Sonnet 5 $0.00009 $0.00435
Haiku 4.5 $0.00005 $0.00218

Measured 9d ago against content hash d2f0f058a8c1, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

forge-mcp-clients scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/forge-mcp-clients/SKILL.md · 119 lines

How it starts

The opening of the file, as written. The whole thing — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Forge playbook — MCP-as-client doctrine (dormant, opt-in)

Do not duplicate the config — defer to: config/orchestration/mcp-registry.json (the server catalog) and config/orchestration/mcp-grants.json (the per-Boss least-privilege matrix) as the single source of truth for tiers, servers, and grants. This file is doctrine/orchestration only — it explains how to use those configs correctly, it does not re-list every server or re-derive the matrix.

Hard rules

  • Nothing here installs, connects to, or activates any MCP server. The zero-dependency Forge default (fs/path only, no live network client) stays 100% intact until the owner explicitly opts a server in.
  • A server is usable only when both are true: (1) the owner has opted it in (a marker naming the server id — see "Opt-in flow" below), and (2) the requesting Boss's grant covers it (tier + server id, per mcp-grants.json). Either condition failing means dormant → honest native fallback.
  • Tier 3 (WRITE-PRIMITIVE) is never auto-used. Every tier-3 call routes through the existing forge-actiongate.cjs hard-gate classifier (the same one used for deploy/push/spend/DNS/etc.) — it is never bypassed and never reimplemented in parallel.
  • Never claim an MCP tool ran, was available, or was opted in when it wasn't. Absence is reported as a labeled native fallback, not silently skipped.

What "MCP-as-client" means here

Forge already ships one MCP artifact: forge-bin/forge-mcp.cjs, a read-only server that exposes this project's own Forge run state (runs/events/reports/memory) to an external MCP host. That is Forge acting as a server — unrelated to this skill.

This skill is the opposite direction: Forge (its Bosses) acting as an MCP client, consuming external MCP servers (docs lookup, web search, browser automation, GitHub, etc.) as extra tools during a task. The client side is deliberately dormant by default — a catalog and a validator, not a running connection.

Read the full file on GitHub · 119 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 9d ago First seen · 119 lines · 46 tokens per session scan A d2f0f058a8c1

Subscribe to this mod's changes

forge-mcp-clients is a skill published in the GitHub repository ForgeyClap/claude-forge (2 stars, last pushed 1mo ago), licensed MIT. It adds 46 tokens to every session and 2,176 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.