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/cloudshipai/station/claude-mdgit clone --depth 1 https://github.com/cloudshipai/stationWhat 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 | $0.08364 | $0.08364 |
| Opus 5 | $0.04182 | $0.04182 |
| Sonnet 5 | $0.01673 | $0.01673 |
| Haiku 4.5 | $0.00836 | $0.00836 |
Grade C, and why
station CLAUDE.md scanned grade C with 3 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.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl -sSL https://install.station.dev | bash Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
RUN apt-get update && apt-get install -y curl wget netcat Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
os.system(f"tar -czf {filename} /data/") # COMMAND INJECTION! How it starts
The opening of the file, as written. The whole thing — 894 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Claude Agent Context for Station Development
Project Overview
Station is a secure, self-hosted platform for creating intelligent multi-environment MCP agents that integrate with Claude. The system allows users to build AI agents that understand development workflows, manage sensitive tools across environments, and run automated background tasks.
Current System State
⚠️ Known Issues
GenKit v1.2.0 Incompatibility
- Status: Station pinned to GenKit v1.1.0 (working)
- Issue: GenKit v1.2.0+ breaks tool registration with child registry pattern
- Impact: "tool not found" errors on both OpenAI and Gemini
- Root Cause: PR #3753 changed tools to register in child registry instead of parent
- Workaround: Stay on v1.1.0 (stable, tested, works perfectly)
- Documentation: See
docs/GENKIT_V1.2.0_INCOMPATIBILITY.mdfor full details - Future Action: Need to adapt tool registration pattern or report upstream
✅ Completed Major Architecture Overhaul
-
Modular Handler Architecture: ✅ Complete - Split 5 large files (5,777 lines) into 43 focused modules
- All handler modules now under 500 lines for maximum maintainability
- Clean separation of concerns with
agent/,file_config/,load/,mcp/,webhooks/modules - Shared utilities in
common/module for DRY code organization
-
File-Based Configuration System: ✅ Fully Migrated from database-based to file-based MCP configs
- GitOps-ready configuration management with Go template support
- Template variable resolution for environment-specific deployments
- Removed old
mcp_configs,template_variables, and related database tables - Updated all services, handlers, and TUI components to use file-based system
-
CLI Agent Execution: ✅ Implemented with clean server/fallback architecture
- Server Mode: Uses
POST /api/v1/agents/:id/queuewith ExecutionQueueService for full execution - Fallback Mode: Provides simplified execution when server not available
- Clean DRY: Reuses existing execution architecture, no code duplication
- Webhook Support: Full webhook notifications work in server mode
- Graceful Degradation: Clear messaging about mode limitations
- Server Mode: Uses
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 · 894 lines · 8,364 tokens per session scan C 439a6fbe19ec
station CLAUDE.md is an instructions file published in the GitHub repository cloudshipai/station (429 stars, last pushed 7mo ago), licensed Apache-2.0. It adds 8,364 tokens to every session, about $0.0418 per session on Opus 5. A static security scan graded it C with 3 findings (downloads and executes remote code, makes network calls, runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
ken CLAUDE.md
Instructions for townsendmerino/ken, covering claude.md, what this is, repository ownership (read this first), commands and embedding parity & golden fixtures (now in aikit).
mosaic AGENTS.md
Instructions for hexxla/mosaic, covering project instructions, your responsibilities as an agent, hexagonal architecture, core concepts and layer responsibilities & rules.
symbolpeek-mcp CLAUDE.md
Instructions for pioner92/symbolpeek-mcp: SymbolPeek's whole value is context economy: an agent gets one symbol, not the file. The rule applies to the code too — including this file, which loads into every session, so hold it to its own bar.
rocketship AGENTS.md
Instructions for rocketship-ai/rocketship, covering agents.md, rocketship cloud v1 snapshot, architecture overview, development commands and build and install.
rocketship CLAUDE.md
Instructions for rocketship-ai/rocketship, covering claude.md, rocketship cloud v1 snapshot (for agents), architecture overview, development commands and build and install.
leather AGENTS.md
AGENTS.md instructions for TGPSKI/leather, covering agents.md, subagent routing table, what leather is, core capabilities and vocabulary.