Skip to content
Agentmart
Submit a tool

agent-message

Inspect the permitted family roster, send durable direct family messages, or create and resolve root-owned human attention requests.

Skill by clankerwrangler

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/clankerwrangler-agent-message

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Inspect the permitted family roster, send durable direct family messages, or create and resolve root-owned human attention requests.

Use cases: agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Python package

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/agent-message

What's inside

Skill name
agent-message
Description
Inspect the permitted family roster, send durable direct family messages, or create and resolve root-owned human attention requests.
License
MIT
Instructions
30 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts

Files (3)

  • SKILL.md
  • pyproject.toml
  • __init__.py · src/harness_agent_message/

SKILL.md (first lines)

# Agent Message

Import with `import harness_agent_message as agent_message` (the kernel also preloads `agent_message`).

Use `agent_message.list_agents()` to inspect reachable agents. Reach is strictly local: an agent may address only its parent, direct children, and same-parent siblings. Grandparents, grandchildren, cousins, and other transitive relationships are rejected even when an exact stable ID is supplied.

Use `agent_message.send(target, body)` for direct delivery. A busy target receives the message at the next safe input/tool checkpoint without interrupting a running tool. An idle target can start a turn. The operation commits synchronously and returns an awaitable `SkillResult`, so both `agent_message.send(...)` and `await agent_message.send(...)` are safe and perform exactly one host call. There is no broadcast operation, and sender identity always comes from the harness connection.

Messages never produce an automatic reply. Send only an intentional, useful result or request; do not echo acknowledgements or respond solely because another agent sent a message.

## Starting a root session (Home only)

`agent_message.start_session(name, task, cwd)` opens a new root session named `name` in the absolute
folder `cwd` and sends it `task` as a message from you, so it can reply to you. Only Home may
call it; other sessions are refused.

## Human attention (root only)

When the human must supply a decision, approval, or missing information, the root can use
`agent_message.request_attention(key, title, body, expires_in=86400)` alongside its normal
conversation explanation. Use explicit, concise plain text (title ≤120 characters,
body ≤1000) and a stable key (≤128 letters, digits, `.`, `_`, `:`, `-`). The expiry is
1–604800 seconds. Children send the need to their parent; they cannot address the human
through this API, and ordinary family messaging rules are unchanged.

Reuse the same key when retrying the same request. Creation is durable and idempotent:
retries do not reopen resolved/expired requests or extend deadlines. A genuinely new
request needs a new key. Call `agent_message.resolve_attention(key)` when that specific
need no longer requires human action. Read acknowledgement is not resolution; expiry
means expired, not approved or completed. Notifications never grant permission, and
creating one does not substitute for an interactive approval's live response mechanism.

Routine progress and every-message alerts do not belong here. Family-idle status and
explicitly intended cron deliveries are tracked separately by the supervisor.

From clankerwrangler/persistent-harness, MIT.

More from clankerwrangler/persistent-harness

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 16, 2026 (13 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from clankerwrangler

Cite this page

agent-message skill. Agent Mart. https://agentmart.dev/skills/clankerwrangler-agent-message (last updated 2026-10-24)