ghostlink CLAUDE.md

A project instruction file for Claude Code, an AI coding assistant, working on GhostLink. GhostLink is a local MCP server that gives coding agents controlled access to repository and Git operations on macOS.

In plain words
What is it for?
Use it when developing or reviewing GhostLink. It explains how to install, build, test, lint, type-check, and run the project, along with where its main source code is organized.
Why use it?
It gives the assistant the project's commands, architecture, design patterns, and required verification checks. This reduces the chance of making changes that do not fit the repository or pass its checks.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/bgorzelic/ghostlink/claude-md
Clone the repo
git clone --depth 1 https://github.com/bgorzelic/ghostlink
Per session 1,290 This file is loaded in full into every session.
When invoked 1,290 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
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 $0.01290 $0.01290
Opus 5 $0.00645 $0.00645
Sonnet 5 $0.00258 $0.00258
Haiku 4.5 $0.00129 $0.00129

Measured 2d ago against content hash c31ac6a0c8b8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ghostlink CLAUDE.md 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 2d 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.md · 117 lines

How it starts

The opening of the file, as written. The whole thing — 117 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project

GhostLink is a local-first MCP server (STDIO) for agentic coding workflows on macOS. It exposes deterministic, policy-gated tools (repo + git + curated commands) via structured JSON envelopes.

Current phase: Phase 1 -- Local STDIO (shipped as v0.1.0). Phase 2 (remote transport) and Phase 3 (agent runtime) are designed but not yet built.

Commands

npm install          # install dependencies
npm run dev          # dev server with watch mode
npm run build        # compile TypeScript to dist/
npm start            # run compiled server (STDIO)
npm test             # run full test suite (vitest, 108 tests)
npm run lint         # ESLint
npm run typecheck    # tsc --noEmit

Single test file: npx vitest run tests/core/tools/repo-search.test.ts Single test by name: npx vitest run -t "returns deterministic envelope"

Verification gate (run after every code change):

npm test && npm run lint && npm run typecheck && npm run build

Architecture

src/
  index.ts                     # Entry point (STDIO transport binding)
  server.ts                    # MCP server factory + tool registration (transport-agnostic)
  core/
    tools/                     # Tool implementations (one file per tool)
      repo-search.ts           # ripgrep-powered search
      repo-read-file.ts        # File read with size/binary guards
      repo-apply-patch.ts      # Pure JS unified diff with atomic rollback
      repo-run.ts              # Curated command runner (allowlist only)
      git-status.ts            # Normalized git status parser
      git-diff.ts              # Diff with sandbox validation + output caps
    policy/
      sandbox.ts               # Path canonicalization, traversal rejection, symlink escape detection
      commands.ts              # Command allowlist registry with per-command arg validation
    utils/
      envelope.ts              # ToolEnvelope builder (ok/error + provenance)
      process.ts               # Safe process spawning (shell:false only)
    logging/
      audit.ts                 # JSONL audit logger (stdout/file/off modes)
  types/
    tool-envelope.ts           # ToolEnvelope<T> type definition
tests/                         # Mirrors src/ structure (vitest)
docs/                          # Public documentation
templates/                     # Ready-to-use config templates for target projects

Read the full file on GitHub · 117 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. 2d ago First seen · 117 lines · 1,290 tokens per session scan A c31ac6a0c8b8

Subscribe to this mod's changes

ghostlink CLAUDE.md is an instructions file published in the GitHub repository bgorzelic/ghostlink (0 stars, last pushed 27d ago), licensed ISC. It adds 1,290 tokens to every session, about $0.0064 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.