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/kornia/bubbaloop/claude-mdgit clone --depth 1 https://github.com/kornia/bubbaloopWhat 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 | $0.04789 | $0.04789 |
| Opus 5 | $0.02395 | $0.02395 |
| Sonnet 5 | $0.00958 | $0.00958 |
| Haiku 4.5 | $0.00479 | $0.00479 |
Grade B, and why
bubbaloop CLAUDE.md scanned grade B with 1 finding against 26 rules in 11 categories β prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency β measured 3d 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
- mold linker config ready in `.cargo/config.toml` β activate with `sudo apt install mold clang` How it starts
The opening of the file, as written. The whole thing β 243 lines β stays where its author put it; the contents beside it link to each section on GitHub.
π¦ Bubbaloop
Physical AI orchestration built on Zenoh. Single binary: CLI + daemon + MCP server.
Living Documents (update these as architecture evolves)
- ARCHITECTURE.md β Layer model, node contract, security, open-core boundary
- ROADMAP.md β Implementation phases (YAML skills, agent, SQLite memory, scheduling)
- CONTRIBUTING.md β Agentic workflows, agent tiers, validation, two-critic loop
- docs/concepts/dataflow.md β Provenance envelope, manifest queryable,
bubbaloop dataflowCLI + MCP tool (how the SDK answers "what's wired, what's flowing")
Structure
crates/bubbaloop/ # Main binary (CLI + daemon + MCP server)
crates/bubbaloop-node/ # Node SDK (standalone, NOT in workspace β batteries-included framework)
crates/bubbaloop-node-build/ # Build helper for nodes (wraps prost-build, standalone)
crates/bubbaloop-schemas/ # Protobuf schemas (standalone, NOT in workspace β never add to workspace)
dashboard/ # React + Vite + TypeScript
Key source files in crates/bubbaloop/src/:
cli/login.rsβ login/logout/status +has_claude_credentials()(env var β OAuth β key file)cli/agent_setup.rsβbubbaloop agent setup: interactive provider/model/identity wizard (no daemon needed)cli/agent_client.rsβ Thin Zenoh CLI client for agent chat/list (pub/sub, no LLM)agent/mod.rsβ Agent core: EventSink trait, run_agent_turn(), AgentTurnInput (soul_path triggers onboarding)agent/gateway.rsβ Zenoh gateway wire format (AgentMessage, AgentEvent, topic builders)agent/runtime.rsβ Multi-agent runtime: AgentsConfig, AgentRuntime, ZenohSink, agent_loop;agent_directory()(pub)agent/prompt.rsβ System prompt builder;build_system_prompt_with_soul_path()injects onboarding prompt for new agentsagent/soul.rsβ Soul struct, first-run onboarding, notify hot-reload (~/.bubbaloop/soul/)agent/provider/mod.rsβ ModelProvider trait, Message, ContentBlock (incl.Image), ToolDefinition, StreamEventagent/provider/gemini.rsβ Gemini API client (recommended default β free tier, vision-capable)agent/provider/claude.rsβ Claude API client with dual auth (API key + OAuth bearer token)agent/provider/ollama.rsβ Ollama local LLM client with tool calling (/api/chat)agent/camera.rsβgrab_frameimplementation: Zenoh query for one JPEG frame + JSON metadata attachment, emitsContentBlock::Imageagent/gateway.rsβ Zenoh + HTTP+SSE agent gateway; per-agent cancel topicbubbaloop/global/{machine_id}/agent/{agent_id}/cancelagent/memory/β 4-tier: world state (live SQLite) + short-term (RAM) + episodic (NDJSON) + semantic (SQLite)agent/heartbeat.rsβ Adaptive heartbeat: arousal + decay + state collectionagent/dispatch.rsβ Internal tool dispatch (47 public MCP tools + ~43 agent-internal tools incl.grab_frame, telemetry, beliefs, constraints, missions)cli/node/mod.rsβ node CRUD, validation, list/add/removecli/node/install.rsβ install, precompiled binary download, GitHub clonecli/node/lifecycle.rsβ start, stop, restart, logscli/node/build.rsβ build nodedaemon/mod.rsβ skill runtime: registry + lifecycle + health + MCP + telemetry watchdogdaemon/telemetry/β Resource watchdog: sampler (sysinfo), circuit breaker, SQLite storage, hot-reload configdaemon/node_manager.rs(57KB) β node lifecycle, build queue, healthdaemon/systemd.rs(38KB) β D-Bus/zbus integrationdaemon/registry.rsβ~/.bubbaloop/nodes.jsonmanagementregistry.rsβ marketplace fetch/parse/cache,find_curl()marketplace.rsβ shared precompiled binary download logic (used by CLI and MCP)mcp/mod.rsβ MCP tools, BubbaLoopMcpServer, ServerHandler implmcp/platform.rsβ PlatformOperations trait, DaemonPlatform, MockPlatform
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.
- 3d ago First seen Β· 243 lines Β· 4,789 tokens per session scan B 6f9da005c53f
bubbaloop CLAUDE.md is an instructions file published in the GitHub repository kornia/bubbaloop (28 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 4,789 tokens to every session, about $0.0239 per session on Opus 5. A static security scan graded it B with 1 finding (asks for root). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
VSLAM-LAB CLAUDE.md
Claude Code instructions for VSLAM-LAB/VSLAM-LAB, covering claude.md, what this is, environment and commands, project structure and adding a baseline or dataset.
pixeltable CLAUDE.md
Claude Code instructions for pixeltable/pixeltable, covering claude.md, protected configuration, project overview, repository structure and setup commands.
mobilegym AGENTS.md
Instructions for Purewhiter/mobilegym, covering agents.md, project overview, type-checking strategy, eslint and with data graph generation.
autonomous-os AGENTS.md
Instructions for autonomous-ai/autonomous-os, covering agents.md, multi-ide rules, working style, parallel work / subagents and device access rules.
albucore AGENTS.md
Instructions for albumentations-team/albucore, covering codex ai development guidelines for albucore, project overview, core documentation, repo-local codex skills and development principles.
screenpipe CLAUDE.md
Claude Code instructions for screenpipe/screenpipe: Conventions live in one file so they cannot drift. Read AGENTS.md.