Getting it into your agent
There is no command for this one: it runs only inside a plugin, and the catalogue could not identify which plugin ships it. The source is linked below.
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.
[](https://agentmods.dev/instructions/mtzanidakis/praktor/claude-md)<a href="https://agentmods.dev/instructions/mtzanidakis/praktor/claude-md"><img src="https://agentmods.dev/badge/instructions/mtzanidakis/praktor/claude-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.06792 | $0.06792 |
| Opus 5 | $0.03396 | $0.03396 |
| Sonnet 5 | $0.01358 | $0.01358 |
| Haiku 4.5 | $0.00679 | $0.00679 |
Grade A, and why
praktor 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 6d 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 — 362 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Praktor
Personal AI Agent Assistant.
Do NOT commit or push unless explicitly asked.
Quick Context
Go 1.26 service (github.com/mtzanidakis/praktor) that connects to Telegram, routes messages to named agents running Claude Code (Agent SDK) in isolated Docker containers, and provides a Mission Control Web UI. Single binary deployment via Docker Compose.
Architecture
Telegram ←→ Go Gateway ←→ Router ←→ Embedded NATS ←→ Agent Containers (Docker)
↕ ↕
SQLite DB Claude Code SDK
↕
Web UI (React SPA)
The gateway binary runs all core services: Telegram bot, message router, NATS message bus, agent orchestrator, scheduler, swarm coordinator, and HTTP/WebSocket server. Agent containers are spawned on demand via the Docker API and communicate with the host over NATS pub/sub.
Named Agents: Multiple agents are defined in YAML config, each with its own description, model, image, env vars, secrets, allowed tools, and workspace. Messages are routed to agents via a 3-tier chain: @agent_name prefix → smart routing via default agent container → default agent fallback.
Project Structure
cmd/praktor/main.go # CLI: `gateway`, `vault`, `backup`, `restore`, and `version` subcommands
cmd/ptask/main.go # Task management CLI (Go, runs inside agent containers)
internal/
config/ # YAML config + env var overrides
extensions/ # Agent extension types (MCP servers, plugins, skills)
store/ # SQLite (modernc.org/sqlite, pure Go) - agents, messages, tasks, swarms, secrets
vault/ # AES-256-GCM encryption with Argon2id key derivation
natsbus/ # Embedded NATS server + client helpers + topic naming
container/ # Docker container lifecycle, image building, volume mounts
agent/ # Message orchestrator, per-agent queue, session tracking
agentmail/ # AgentMail WebSocket client for real-time email events
speech/ # OpenAI Speech API client (Whisper STT + TTS)
registry/ # Agent registry - syncs YAML config to DB, resolves agent config
router/ # Message router - @prefix parsing, smart routing via default agent
telegram/ # Telegram bot (telego), long-polling, message chunking
scheduler/ # Cron/interval/relative delay task polling (adhocore/gronx)
swarm/ # Graph-based swarm orchestration (DAG execution, collaborative chat)
web/ # HTTP server, REST API, WebSocket hub, embedded SPA
Dockerfile # Gateway image (multi-stage: UI + Go + scratch)
Dockerfile.agent # Agent image (multi-stage: Go + playwright-cli + esbuild + alpine)
agent-runner/src/ # TypeScript: NATS bridge + Claude Code SDK + MCP servers (bundled with esbuild)
index.ts # Main entrypoint: agent lifecycle, message handling (sequential user messages + parallel scheduled tasks), MCP server registration
extensions.ts # Apply agent extensions on startup (MCP servers, plugins, skills)
nats-bridge.ts # NATS pub/sub wrapper for agent ↔ host communication
ipc.ts # Shared NATS IPC helper (sendIPC + IPCResponse)
mcp-tasks.ts # MCP server: scheduled_task_create/list/delete
mcp-profile.ts # MCP server: user_profile_read/update
mcp-memory.ts # MCP server: memory_store/recall/list/delete/forget + vector embeddings
mcp-swarm.ts # MCP server: swarm_chat_send (conditional on SWARM_CHAT_TOPIC)
mcp-nix.ts # MCP server: nix_search/add/list_installed/remove/upgrade
mcp-file.ts # MCP server: file_send (send files to Telegram)
ui/ # React/Vite SPA (dark theme, indigo accent)
src/pages/ # Dashboard, Agents, Conversations, Tasks, Secrets, Swarms
src/components/Login.tsx # Session-based login form
src/components/SwarmGraph.tsx # SVG-based visual graph editor for swarm topology
src/hooks/useWebSocket.ts # Real-time WebSocket event hook
config/praktor.example.yaml # Example configuration
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.
- 6d ago First seen · 362 lines · 6,792 tokens per session scan A 877652282682
praktor CLAUDE.md is an instructions file published in the GitHub repository mtzanidakis/praktor (42 stars, last pushed yesterday), licensed MIT. It adds 6,792 tokens to every session, about $0.0340 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 2026-08-30.
Other instructions, from other repositories
homebutler CLAUDE.md
Claude Code instructions for Higangssh/homebutler, covering homebutler, verify before claiming anything works, writing commits, prs, and comments, no tool attribution and commit subjects.
nora CLAUDE.md
Claude Code instructions for solomon2773/nora, covering claude.md, what this repo is, development commands, docker compose (primary path) and per-service dev (without docker).
hrdx AGENTS.md
AGENTS.md instructions for patriceckhart/hrdx, covering working agreement for hrdx, product intent, starting a task, code ownership map and correctness contracts.
dense-mem AGENTS.md
AGENTS.md instructions for markhuangai/dense-mem, covering dense-mem repository guidance, project context, architecture decision records, current stack and target architecture.
piper CLAUDE.md
Claude Code instructions for piperbox/piper, covering claude.md, what this is, coding principles, 1. think before coding and 2. simplicity first.
piper AGENTS.md
AGENTS.md instructions for piperbox/piper: Read CLAUDE.md in full first — it is the authoritative source for coding principles, commands, constraints, layering, and branch/PR/issue workflow. This file only captures high-signal facts an agent would otherwise get wrong that aren't obvious from filenames or CLAUDE.md.