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 skills/oryntai/agentlink/agentlink-messengernpx skills add Oryntai/AgentLink --skill agentlink-messengergit clone --depth 1 https://github.com/Oryntai/AgentLinkWrote 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/oryntai/agentlink/agentlink-messenger)<a href="https://agentmods.dev/skills/oryntai/agentlink/agentlink-messenger"><img src="https://agentmods.dev/badge/skills/oryntai/agentlink/agentlink-messenger.svg" alt="Measured on agentmods" 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 | $0.00053 | $0.01013 |
| Opus 5 | $0.00026 | $0.00507 |
| Sonnet 5 | $0.00011 | $0.00203 |
| Haiku 4.5 | $0.00005 | $0.00101 |
Grade A, and why
agentlink-messenger 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 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.
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 — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AgentLink Messenger
Use the AgentLink MCP tools as part of normal collaboration, not as a one-off transcript dump. The peer has no direct access to this machine; every reply remains this agent's judgment.
Routine
- At task start, call
peer_status,peer_inbox_list, andpeer_request_statusas appropriate. Repeat after substantial work, after an owner notification, and before the final response. Do not poll tightly. - Answer before asking. Before every
peer_askorpeer_request_send, checkpeer_inbox_listand deal with what is already waiting. When both agents write at the same time the conversations cross and both lose a round; reading first prevents it. Heed the unread-inbound warning these tools return. - Pick the right message type.
peer_askwhen the answer blocks current work.peer_request_sendwhen local work can continue; retain the request ID and collect it later withpeer_request_status.peer_notifyfor one-way information that needs no reply — progress updates, "the server is up", "wrapping up". Never send an FYI as a request, and never answer a received FYI. - A request you sent stays an obligation on the peer until it is answered. If a later message supersedes it, or you sent it by mistake, withdraw it with
peer_cancel— telling the peer in text to "ignore my last message" leaves an unanswered request in its inbox, and it will dutifully answer both. Withdrawal only works until the peer's agent claims it; after that, send a correction instead. - When serving a peer, use
peer_listenor claim an item withpeer_inbox_claim, inspect only the necessary local resources, then answer withpeer_respond. A dedicated responder task is a live service: keeppeer_listenpending withtimeout_seconds: 86400; after everypeer_respond, timeout, FYI, or recoverable error, check the inbox and immediately re-armpeer_listen. - For a multi-step joint plan, negotiate it with
peer_goalbefore usingpeer_exchangeorpeer_reply; finish it withpeer_complete. Completion does not end the link. - Never leave silently. Before ending your part of a collaboration: answer everything still queued, then propose
peer_completeor send a finalpeer_notify. Keep shared services (dev servers, databases the peer tests against) running until the peer confirms it no longer needs them.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 43 lines · 53 tokens per session scan A 2851e662d7d7
agentlink-messenger is a skill published in the GitHub repository Oryntai/AgentLink (0 stars, last pushed 4d ago), licensed MIT. It adds 53 tokens to every session and 1,013 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 2026-08-31.
Other skills, from other repositories
agent-link
Talk to other coding agents — another Claude Code, or a colleague's Codex — over an encrypted room. Use when the user wants to coordinate, split, hand off or divide work with another agent; asks what the other agent is doing; says "chiedi al collega", "dillo all'altro Claude", "dividiamoci il lavoro", "pair with"…
ironmesh-peers
Drill into per-peer IronMesh metrics — retry reasons, session rekeys, offline-since, transport type. Use when ironmesh-status shows a problem on a specific peer and you need detail.
ironmesh-send
Send an application message (MSG) to a specific IronMesh peer. Use when the user asks to "send X to agent Y" or "ping peer Z". Has side effects — only invoke when the user explicitly asks.
ironmesh-audit
Tail the tamper-evident IronMesh audit log for security events. Use when investigating a TOFU mismatch, auth failure, peer dropping out, or a suspected attack.
ironmesh-status
Report IronMesh mesh health — uptime, active peer count, message lifetime percentiles, queue pressure. Use when the user asks "is the mesh ok", "how's ironmesh", or before taking actions that depend on a peer being reachable.
ironmesh-trust
Inspect and manage IronMesh's TOFU trust store — list pinned peers, show revocations, revoke a compromised peer. Revoke has side effects and requires the user's explicit confirmation.