Agent Messaging Protocol
ASecurity- Communicate directly with the intended party whenever possible. - Treat the original message as the authoritative source. - Forward a pointer to the original message instead of retyping or paraphrasing its body. - Include the original author, original message ID, intended recipient, and requested action. - Attach `message:<original-message-id>` as a message reference when supported. - Read the original message before acting: `thrum message get <message-id> --json`. - Reply against the or
- 2 stars
- 0 votes
- 0 copies
- 0 views
- Added November 7, 2026
Security analysis
100/100Pro scans all 2 files and shows the line behind each finding
npx -y skills add thrum-team/thrum-pro --skill agent-messaging-protocol --agent claude-codeAre you the author of Agent Messaging Protocol?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/thrum-team-agent-messaging-protocol-thrum-pro)---
name: agent-messaging-protocol
description:
"Use when forwarding or relaying information between parties, routing an owner
answer, carrying a decision or request through an intermediary, summarizing
another author's message, or adding interpretation to someone else's words."
# source: claude-plugin/skills/agent-messaging-protocol/SKILL.md
---
## Agent Messaging Protocol for Forwarding and Relaying
- Communicate directly with the intended party whenever possible.
- Treat the original message as the authoritative source.
- Forward a pointer to the original message instead of retyping or paraphrasing
its body.
- Include the original author, original message ID, intended recipient, and
requested action.
- Attach `message:<original-message-id>` as a message reference when supported.
- Read the original message before acting:
`thrum message get <message-id> --json`.
- Reply against the original message: `thrum reply <message-id> ...`.
- Do not route the response back through the intermediary.
- Add commentary or interpretation only when requested or necessary.
- Label added reasoning as `INTERPRETATION - @agent`.
- Never present an interpretation, summary, or recommendation as words or
authority from the original author.
- Label a requested summary as `SUMMARY - @agent` and retain the original
message ID.
- Preserve the original scope, urgency, constraints, and requested action.
- Do not introduce new requirements, decisions, risks, or authorization while
routing.
- Ask the original author directly when clarification is required.
- Continue referencing the original message rather than an intermediary's
summary.
- Identify corrections by the exact message they supersede.
- Distinguish delivery, acknowledgement, execution, and verification.
- Do not reply to a message that only acknowledges or thanks; mark it read.
- Preserve the original thread audience when replying unless a deliberately
narrower audience is required.
### Routing Envelope
```text
ROUTE POINTER
From: @original-author
Message: msg_...
Action: Read and reply to the original message.
```
Files in this skill
- SKILL.md
- agents/openai.yaml
Attribution
Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.