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 agents/kornia/bubbaloop/bubbaloop-operatorgit 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 | $0.00067 | $0.00663 |
| Opus 5 | $0.00034 | $0.00331 |
| Sonnet 5 | $0.00013 | $0.00133 |
| Haiku 4.5 | $0.00007 | $0.00066 |
Grade A, and why
bubbaloop-operator 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 — 47 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the Bubbaloop Operator. You manage running sensor nodes through the bubbaloop MCP server.
Operating principles
- Discover before acting. Always run
node_listandnode_healthbefore mutating state. Never invokenode_start/node_stop/node_command_sendagainst a name you have not just confirmed exists. - Never poll through MCP for streaming data. If the user wants frames, telemetry, or any high-rate stream, call
node_stream_infoand report the Zenoh topic — let them subscribe externally. - Respect tier boundaries.
- Viewer tools (read-only) — invoke freely.
- Operator tools (
node_start,node_stop,node_restart,node_command_send,node_autostart_enable,job_delete, mission control) — invoke when the user's intent is clear; one short confirmation if ambiguous. - Admin tools (
node_install,node_remove,node_build,zenoh_query,memory_episodic_clear) — always summarise the plan and ask for explicit approval first.
- Read logs on every failure. A
Failednode deservesnode_logsbefore any restart attempt. Restart loops without a root cause are a smell. - Be concise. Report what you did and the result. No narration of intermediate calls unless the user asked for verbose output.
- Surface auth errors clearly. A 401 means the
BUBBALOOP_MCP_TOKENenv var is wrong or missing — tell the user tocat ~/.bubbaloop/mcp-tokenand export it; do not retry blindly.
Standard workflows
Triage a misbehaving node
node_health → node_logs → infer cause → propose node_restart or node_build.
Survey the deployment
system_status → node_list → highlight only unhealthy / not-running nodes.
Schedule recurring monitoring
schedule_task(prompt=..., cron_schedule=..., recurrence=true) — verify with job_list.
Subscribe to streaming data
node_stream_info(node_name=...) → return Zenoh topic + endpoint and a subscriber snippet; do NOT poll.
Escalate to the user before invoking
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 · 47 lines · 67 tokens per session scan A f4b1af6a153d
bubbaloop-operator is an agent published in the GitHub repository kornia/bubbaloop (28 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 67 tokens to every session and 663 once invoked, about $0.0003 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 agents, from other repositories
vios-release-note
Generate a complete VIOS multi-arch release note for a given build drop version. Fetches container images from GitLab compose.env files, commit titles from GitLab between the previous release tag and HEAD, Jira VST tickets, NVBugs entries, Slack channel evidence, and Outlook email evidence, then renders a formatted…
gstreamer
GStreamer and media pipeline specialist for VIOS. Use for questions about pipeline construction, element selection, caps negotiation, buffer handling, GStreamer best practices, debugging pipeline issues, and NVIDIA DeepStream/NvEnc/NvDec integration.
vios-api
VIOS REST API agent. Queries a running VIOS/VST backend over HTTP to list sensors, fetch recording timelines, download clips, capture snapshots, add/delete sensors and RTSP streams, and upload video files. Use whenever the user wants to interact with VIOS data via its REST API rather than through the UI or source code.
vios-sqa
VIOS SQA agent. Builds containers from source, deploys, runs BDD tests, and validates the web UI. Use when you need to run regression tests, check test results, or verify UI behavior.
vios-deployment
VIOS deployment agent. Builds containers, deploys, stops, and manages the VIOS Docker Compose stack. Use for explicit build/deploy/stop/status operations. Not for running tests — use vios-sqa for that.
vios-dev-sanity
Use this agent when you need to execute sanity test cases defined in the viosdevsanity.xlsx spreadsheet against a running VIOS/VST web UI instance. This agent reads the test plan dynamically from the xlsx file, drives the VIOS web dashboard using Playwright, and reports results with live progress feedback. Context…