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.
npx agentmods add instructions/bgorzelic/ghostlink/agents-mdgit clone --depth 1 https://github.com/bgorzelic/ghostlinkWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00461 | $0.00461 |
| Opus 5 | $0.00230 | $0.00230 |
| Sonnet 5 | $0.00092 | $0.00092 |
| Haiku 4.5 | $0.00046 | $0.00046 |
Grade A, and why
ghostlink AGENTS.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.
How it starts
The opening of the file, as written. The whole thing — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.
GhostLink -- AGENTS.md
Operating Model
GhostLink development runs as a structured agentic swarm with a single Orchestrator and scoped sub-agents.
- Orchestrator: Primary Claude Code instance. Owns plan, milestones, and final verification.
- Sub-agents: Spawned only when isolation or parallelism improves output. Each has a declared scope and must not modify files outside it.
Global Rules
Every agent (orchestrator or sub-agent) must:
- Operate only within the repo root
- Return structured outputs (diffs + command outputs + assumptions)
- Run verification after edits (
npm test && npm run lint && npm run typecheck && npm run build) - Avoid broad refactors unless explicitly requested
- Never introduce arbitrary command execution surfaces
Roles
| Role | Scope | Responsibility |
|---|---|---|
| Orchestrator | Entire repo | Plan, scope, milestones, merge, verify |
| Protocol Engineer | src/server.ts, src/index.ts, src/types/* |
Server factory, tool registration, transport |
| Toolsmith | src/core/tools/*, src/core/utils/*, src/core/policy/* |
Tool implementation, deterministic output |
| Security Reviewer | docs/SECURITY.md, src/core/policy/* |
Threat model, audit, hardening recommendations |
| Test Engineer | tests/* |
Unit + golden tests per tool |
| Docs Engineer | docs/*, root markdown files |
Schema docs, architecture, prompts |
Full sub-agent prompt templates: docs/PROMPTS.md
Handoff Contract
Every sub-agent response must include:
- Unified diff or exact file contents
- Commands executed and their output
- Assumptions and risks identified
- Next actions (if blocked or incomplete)
Long-Term Vision Alignment
All agents must preserve forward-compatibility with:
- Remote transport (Phase 2)
- Memory integration (Phase 3)
- Runtime orchestration (Phase 3)
Never hard-code decisions that block Phase 2 or Phase 3 expansion. GhostLink remains the capability plane. Orchestration lives elsewhere.
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.
- 2d ago First seen · 52 lines · 461 tokens per session scan A 0aed6ff3b146
ghostlink AGENTS.md is an instructions file published in the GitHub repository bgorzelic/ghostlink (0 stars, last pushed 28d ago), licensed ISC. It adds 461 tokens to every session, about $0.0023 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.
Other instructions, from other repositories
virtualization-mcp AGENTS.md
Instructions for sandraschi/virtualization-mcp, covering virtualization-mcp — agent guide, entry points, standards, key files and quick ref.
onlyboxes AGENTS.md
AGENTS.md instructions for Coooolfan/onlyboxes, covering 项目结构, 项目概述 and 注意事项.
onlyboxes CLAUDE.md
Claude Code instructions for Coooolfan/onlyboxes, a project described as: Code Execution Sandbox Platform Solution for Individuals and Small Teams with MCP.
virtualization-mcp CLAUDE.md
Instructions for sandraschi/virtualization-mcp, covering virtualization-mcp — claude code guide, entry points, standards, key files and commands.
brood-box CLAUDE.md
Instructions for stacklok/brood-box, covering brood-box, commands — always use task (taskfile.yaml), architecture — strict ddd (domain-driven design), layers and ddd rules (non-negotiable).
intervals-icu-mcp CLAUDE.md
Instructions for hhopke/intervals-icu-mcp, covering claude.md, project overview, development commands, architecture (quick reference) and tool categories.