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 5dive-ai/5dive-plugins --skill connect-sitegit clone --depth 1 https://github.com/5dive-ai/5dive-pluginsWrote 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/5dive-ai/5dive-plugins/connect-site)<a href="https://agentmods.dev/skills/5dive-ai/5dive-plugins/connect-site"><img src="https://agentmods.dev/badge/skills/5dive-ai/5dive-plugins/connect-site/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/5dive-ai/5dive-plugins/connect-site"><img src="https://agentmods.dev/badge/skills/5dive-ai/5dive-plugins/connect-site.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00119 | $0.03369 |
| Opus 5.5 | $0.00048 | $0.01348 |
| Sonnet 5.5 | $0.00024 | $0.00674 |
| Haiku 4.5 | $0.00012 | $0.00337 |
Grade B, and why
connect-site scanned grade B with 2 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 16d 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.
sudo journalctl -u shelld -n 200 --no-pager \ Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
open it, do not curl it, do not "just check that it works", do not paste it into a fetch How it starts
The opening of the file, as written. The whole thing — 206 lines — stays where its author put it; the contents beside it link to each section on GitHub.
connect-site — use the bound dashboard handoff, then stop and check
You cannot log a person in. What you CAN do is put a real browser on the box, mint a
one-time link to it, hand that link to the human over whatever channel you are paired to,
and wait. The login they perform lands in a profile that outlives the viewer and that
your later run calls reuse.
The commands are easy. The orchestration is where seats get it wrong, and every mistake here is silent — nothing errors, the human just never gets in.
THE RULE THAT COSTS THE MOST WHEN BROKEN
The viewer link is the HUMAN'S, and it is spent by the first successful GET. Do not open it, do not curl it, do not "just check that it works", do not paste it into a fetch tool, do not preview it. A one-time ticket verified by you is a ticket denied to them — they get a dead page and you get a bug report about the wrong thing. (This is not hypothetical: a lead seat did exactly this on 2026-09-13 while debugging.)
To check a link you already handed over, read the box journal, never the link:
sudo journalctl -u shelld -n 200 --no-pager \
| grep -E 'viewer_bind|viewer_redeemed|viewer_denied|viewer_ws_connected'
viewer_bind— the link was minted (yours).viewer_redeemed+viewer_ws_connected— the human is in. Stop worrying.viewer_deniedwith areason— that reason is your whole diagnosis (no live bind,redeem refused, malformed, wrong session).- nothing at all — they have not clicked yet. Wait, or re-mint after the TTL.
The other safe probe is a deliberately bad nonce: a refusal is a pure refusal and spends nothing. A good nonce is the customer's only redemption.
When YOU need the owner logged in: ask with a button (DIVE-4992)
Do not walk the owner through the dashboard, and never hand them a link you made yourself. Ask from your own seat:
5dive browser connect-request <site> --reason="<one line: why you need it>"
Your paired owner gets a Telegram message with a Connect button. Nothing is
bound until they tap it. Their tap opens the browser on the box and sends them the
one-time link as code, from root, not from you: you never see it, so you cannot spend it.
They log in, then tap Done. Done revokes the view, stops the browser and probes, in
that order. You get a channel message at each step, [browser connect] …. After the
Done message, 5dive browser status <site> has the verdict. Do not poll while you wait.
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.
- 16d ago Changed · +14 lines 7ada08590511
- 19d ago Changed · +1 lines 5dd1bf2f97be
- 23d ago Changed · +26 lines · +12 tokens per session dc3e2f2b4783
- 26d ago Changed · +14 lines 5e64fe1d9cbd
- 28d ago Changed · +1 lines a850a10b1911
- 29d ago Changed 63c41c77dd34
- 1mo ago Changed · +20 lines 90e653543075
- 1mo ago Changed · +11 lines 0ac2a7b064e9
- 1mo ago First seen · 119 lines · 107 tokens per session scan B ed2574641d3e
connect-site is a skill published in the GitHub repository 5dive-ai/5dive-plugins (2 stars, last pushed today), licensed Apache-2.0. It adds 119 tokens to every session and 3,369 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it B with 2 findings (asks for root, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-14.
Other skills, from other repositories
notebooklm
Use this skill to query your Google NotebookLM notebooks for source-grounded, citation-backed answers from Gemini. Browser automation, library management, persistent auth.
aident-loadout
Run multi-step work across apps the user connected in Aident Loadout: live action schemas, connection checks, cost preflight, and approval before side effects.
autonomous-agent-patterns
Design patterns for building autonomous coding agents, inspired by Cline and OpenAI Codex.
test-site
Tests a deployed, activated Power Pages site at runtime using browser-based navigation, page crawling, and API request verification via Playwright. Use when the user wants to test, verify, or smoke-test their deployed site.
preview-screens
Use when the user wants to preview generated screens in a browser without starting Metro / a simulator — for example after /create-mobile-app finishes or after /edit-app regenerates a screen.
qa
Browser-based QA verification. Launches a real browser, navigates the app, clicks buttons, fills forms, and tests user flows. Works as a standalone skill or as a phase end condition in campaigns. Requires Playwright (optional dependency, graceful skip if not installed).