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 skills add inkbox-ai/inkbox --skill inkbox-tunnelsgit clone --depth 1 https://github.com/inkbox-ai/inkboxWrote 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/skills/inkbox-ai/inkbox/inkbox-tunnels)<a href="https://agentmods.dev/skills/inkbox-ai/inkbox/inkbox-tunnels"><img src="https://agentmods.dev/badge/skills/inkbox-ai/inkbox/inkbox-tunnels/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/inkbox-ai/inkbox/inkbox-tunnels"><img src="https://agentmods.dev/badge/skills/inkbox-ai/inkbox/inkbox-tunnels.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high System Prompt Leakage · line 325 Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.Fix: Remove any instructions that reveal, print, or output system prompts or internal rules. System instructions should never be exposed to end users.
- medium Rogue Agent · line 64 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
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 | $0.00047 | $0.03830 |
| Opus 5 | $0.00023 | $0.01915 |
| Sonnet 5 | $0.00009 | $0.00766 |
| Haiku 4.5 | $0.00005 | $0.00383 |
Grade A, and why
inkbox-tunnels 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 10d 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 — 384 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Inkbox Tunnels
Inkbox Tunnels expose a process running on the developer's machine (or any POSIX host) at a public https://{handle}.inkboxwire.com URL. The SDK opens an outbound HTTP/2 connection to the data plane; inbound third-party traffic rides back over that same connection. No inbound port to open, no static IP needed.
Tunnels are an identity property: every agent identity owns exactly one tunnel, and tunnel_name == agent_handle. There is no standalone "create tunnel" call — the tunnel is provisioned atomically by inkbox.createIdentity(...). Pre-existing identities created before this rollout already have tunnels (the migration backfilled them).
Two TLS modes:
| Mode | Who terminates TLS | When to pick |
|---|---|---|
edge (default) |
Inkbox terminates at the edge, forwards plaintext requests as envelopes | Most apps. Simpler. The SDK sees parsed Requests / ASGI scopes. |
passthrough |
The SDK terminates TLS in your process | When customers must speak directly to your cert (mTLS, custom SNI), or when end-to-end encryption is a hard requirement. |
Both modes accept either a forward_to URL (proxy to a local HTTP server) or an in-process callable (Fetch handler in TS, ASGI app in Python). WebSocket upgrades work on either.
Platform: tunnels require POSIX. connect() raises on Windows; read-only control-plane calls work everywhere.
Authentication
The data plane authenticates with the same API key the SDK client was constructed with. Two options:
- Admin-scoped key — can connect any tunnel in the org.
- Identity-scoped key — only valid for the tunnel attached to its scoped identity.
For a fresh agent bootstrap, mint an identity-scoped key after createIdentity:
identity = inkbox.create_identity("my-app")
scoped_key = inkbox.api_keys.create(label="my-app agent", scoped_identity_id=identity.id)
# Hand `scoped_key.api_key` to the agent process. Its `Inkbox(api_key=...)`
# will use it for both REST and the tunnel data plane.
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.
- 10d ago First seen · 384 lines · 47 tokens per session scan A 61005e9e26cb
inkbox-tunnels is a skill published in the GitHub repository inkbox-ai/inkbox (78 stars, last pushed today), licensed MIT. It adds 47 tokens to every session and 3,830 once invoked, about $0.0002 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 skills, from other repositories
vs-project
Create Viking web projects, start and verify a local preview, or deploy a generated project to Volcengine IGA Pages when explicitly requested. Includes agent-guided feature, eligible application, dataset, scene, and authentication choices. Use only after confirming the installed CLI exposes vs project; otherwise stop…
qwenpaw-usage
A QwenPaw command-line guide for setting up scheduled self-checks and packaging a QwenPaw workspace for another server.
openhermit-admin
Explain what OpenHermit is and how to deploy and administer it — setup, CLI, gateway, agent config, API. Use when the user asks about installing, operating, or administering an OpenHermit deployment. For how agents behave for their owners/users/guests, see openhermit-guide.
qzcli-mcp
A skill for using qzcli tools to inspect workspaces, resources, jobs, availability, and raw status values on the 启智 platform.
cloudflare
Manage Cloudflare DNS, zones, cache, tunnels, Workers and Pages via the Cloudflare API. Use when the user asks about their domains, DNS records, cache purging, or apps hosted on Cloudflare.
aliyun-ecs
A command-line tool for managing Alibaba Cloud Elastic Compute Service (ECS), which provides rented virtual servers. It handles servers and related resources such as disks, snapshots, images, security groups, keys, and network interfaces.