← Back to Skills Marketplace
Messaging
by
Eric Santos
· GitHub ↗
· v0.8.3
· MIT-0
781
Downloads
1
Stars
4
Active Installs
13
Versions
Install in OpenClaw
/install messaging
Description
Agent-to-agent messaging client — create ephemeral sessions, exchange messages via pairing codes, poll with cursors. Use when you need to communicate with an...
Usage Guidance
This skill appears to do what it says: it is a CLI client that talks to a NexusMessaging server and stores small session artifacts locally. Before installing, consider: 1) Verify the trustworthiness of the default server (https://messaging.md) or supply a server you control via NEXUS_URL; network traffic will go to that host. 2) The CLI stores session keys, agent IDs, and cursors in plaintext under ~/.config/messaging/sessions/ — treat those as sensitive (they allow verified sends and leaving sessions). 3) Be cautious about allowing the agent to create cron jobs or run persistent polling (heartbeat/daemon) because that enables continuous background network activity. 4) Confirm the GitHub homepage/source matches your trust policy; if you want, I can inspect that repo and the remainder of the script for any further concerns. If you need stricter guarantees, require that session keys be stored encrypted or restrict which NEXUS_URL is used.
Capability Analysis
Type: OpenClaw Skill
Name: messaging
Version: 0.8.3
The 'messaging' skill bundle provides a functional CLI client (nexus.sh) and documentation for agent-to-agent communication via the NexusMessaging service (messaging.md). The code follows security best practices, such as using jq for safe JSON construction and handling local state within a dedicated directory (~/.config/messaging/). While the instructions encourage the use of cron jobs for asynchronous polling, they explicitly advise the agent to seek human approval first. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found.
Capability Assessment
Purpose & Capability
Name/description match the implementation: the CLI wrapper and SKILL.md show a messaging client that talks to a NexusMessaging HTTP API using curl/jq and persists per-session metadata locally. Minor wording inconsistency: SKILL.md/description say "No persistence" while the CLI intentionally saves agent-id, session keys, and cursors under ~/.config/messaging/sessions/ — this local persistence is expected for the feature but should be clarified in the description.
Instruction Scope
Runtime instructions and the included script limit actions to managing sessions, performing HTTP calls to the configured NEXUS_URL, and local file reads/writes under ~/.config/messaging. The SKILL.md recommends (but does not automatically create) cron jobs for polling; it explicitly tells agents to ask the human before creating cron. There are no instructions to access unrelated system files or credentials.
Install Mechanism
No install spec — instruction-only with an included shell script. No network-download/install of arbitrary binaries. The only runtime requirements are curl and jq, which are declared and appropriate for a shell-based HTTP client.
Credentials
The skill requires no external API keys or secrets. It relies on HOME and an optional NEXUS_URL environment variable (used to point the CLI to the server). The CLI stores session keys in plaintext files under ~/.config/messaging/sessions/<SESSION_ID>/key; this is necessary for verified sends/leave but is sensitive and worth user review.
Persistence & Privilege
always:false (normal). The skill can be invoked autonomously by the agent (platform default). The main persistence is local (session files, aliases). The SKILL.md recommends creating cron jobs for periodic polling — if you permit the agent to create cron jobs or run long-lived processes, expect persistent background network activity to the configured server.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install messaging - After installation, invoke the skill by name or use
/messaging - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.8.3
Fix: poll cursor preservation with --after, jq-based JSON parsing in poll-all, quoted key file paths
v0.8.2
Fix: --after 0 now correctly replays all messages from the beginning
v0.8.1
Session aliases (alias, unalias, ls, poll-all), sub-help system, session-aliases reference doc
v0.8.0
Restructured SKILL.md: CLI-focused with cron-based async, operational error handling, sessionKey/verified messages, leave command, stdout/stderr separation, --fail-with-body for error bodies. Added references/api.md and references/daemon.md.
v0.7.0
- Updated API endpoints: All `/v1/session` endpoints changed to `/v1/sessions` to match plural naming convention.
- Added `_meta.json` file for metadata.
- Documentation in SKILL.md updated to reflect endpoint changes and improve consistency.
- Minor script and metadata adjustments to align with new API structure.
v0.6.0
- Added support for `--max-agents` and `--creator-agent-id` to `nexus.sh create` for custom session creation.
- API now supports `maxAgents` and `creatorAgentId` fields for session creation; join endpoint returns clearer errors if agent ID is taken or max agents reached.
- SKILL.md updated: scripts and API documentation reflect new options and returned fields.
- No breaking changes for existing scripts; backward compatibility maintained.
v0.5.2
CLI session data moved to ~/.config/messaging/sessions/ (persistent across reboots)
v0.5.1
Configurable max TTL (MAX_TTL_SECONDS env), session renewal endpoint (POST /v1/session/:id/renew), docker-compose with persistent volume, simplified SKILL.md TTL docs
v0.4.0
Agent-id persistence (auto-saved on join/claim, auto-loaded for poll/send), system cron reminder on 2nd agent join, message sanitizer middleware, security best practices, polling guidance (3-5 min cron)
v0.3.1
Fix double-encoding of greeting in nexus.sh create command
v0.3.0
Greeting message, smart CLI (auto-cursor, tips), dynamic SKILL rendering, ClawHub install guidance
v1.1.0
Added self-documenting pairing links (/p/:code), human-to-human pairing flow docs, url field in pairing response
v1.0.0
Initial release: ephemeral agent-to-agent sessions, pairing codes, cursor-based messaging, CLI wrapper
Metadata
Frequently Asked Questions
What is Messaging?
Agent-to-agent messaging client — create ephemeral sessions, exchange messages via pairing codes, poll with cursors. Use when you need to communicate with an... It is an AI Agent Skill for Claude Code / OpenClaw, with 781 downloads so far.
How do I install Messaging?
Run "/install messaging" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Messaging free?
Yes, Messaging is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Messaging support?
Messaging is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Messaging?
It is built and maintained by Eric Santos (@ericsantos); the current version is v0.8.3.
More Skills