feat(prompt): add AI disclosure policy for external service communications by xingyaoww · Pull Request #2476 · OpenHands/software-agent-sdk

xingyaoww · 2026-03-17T06:11:28Z

Summary

Adds an AI disclosure instruction to the <EXTERNAL_SERVICES> section of the system prompt. When the agent posts messages, comments, issues, or any human-readable content to external services, it should include a brief note indicating the content was generated by an AI agent on behalf of the user.

Motivation

Currently, nothing in the system prompt or skills instructs the agent to identify itself as AI when posting to external services. This means messages posted to Slack, GitHub comments, Discord channels, Linear issues, Notion pages, etc. can appear as if they were written directly by a human — which is misleading to readers.

Why the SDK system prompt (not just individual skills)?

The agent can post human-visible messages through many channels beyond the handful of provider skills in the extensions repo:

MCP tools (e.g., Slack, email, CRM integrations)
Custom tools built by SDK users
Raw API calls via curl
Platform-injected tools (e.g., Slack tools in OpenHands Cloud)

Placing the rule in the system prompt ensures universal coverage regardless of the tool or integration used. Individual provider skills in the extensions repo can optionally reinforce with platform-specific formatting guidance.

Changes

openhands-sdk/openhands/sdk/agent/prompts/system_prompt.j2 — Added one bullet point to the <EXTERNAL_SERVICES> section:

AI disclosure: When posting messages, comments, issues, or any content to external services that will be read by humans (e.g., Slack messages, GitHub/GitLab comments, PR/MR descriptions, Discord messages, Linear/Jira issues, Notion pages, emails, etc.), always include a brief note indicating the content was generated by an AI agent on behalf of the user.

The instruction:

Provides a concrete example format
Enumerates common platforms for clarity
Explicitly covers all channels: dedicated tools, MCP integrations, and direct API calls

Impact

All prompt variants (system_prompt_interactive.j2, system_prompt_long_horizon.j2, system_prompt_tech_philosophy.j2) inherit this via {% include "system_prompt.j2" %}
No test changes needed — no Python tests assert on EXTERNAL_SERVICES content
Fixture files containing rendered prompts will naturally update on regeneration

Agent Server images for this PR

• GHCR package: https://github.com/OpenHands/agent-sdk/pkgs/container/agent-server

Variants & Base Images

Variant	Architectures	Base Image	Docs / Tags
java	amd64, arm64	`eclipse-temurin:17-jdk`	Link
python	amd64, arm64	`nikolaik/python-nodejs:python3.13-nodejs22`	Link
golang	amd64, arm64	`golang:1.21-bookworm`	Link

Pull (multi-arch manifest)

# Each variant is a multi-arch manifest supporting both amd64 and arm64
docker pull ghcr.io/openhands/agent-server:bd0530f-python

Run

docker run -it --rm \
  -p 8000:8000 \
  --name agent-server-bd0530f-python \
  ghcr.io/openhands/agent-server:bd0530f-python

All tags pushed for this build

ghcr.io/openhands/agent-server:bd0530f-golang-amd64
ghcr.io/openhands/agent-server:bd0530f-golang_tag_1.21-bookworm-amd64
ghcr.io/openhands/agent-server:bd0530f-golang-arm64
ghcr.io/openhands/agent-server:bd0530f-golang_tag_1.21-bookworm-arm64
ghcr.io/openhands/agent-server:bd0530f-java-amd64
ghcr.io/openhands/agent-server:bd0530f-eclipse-temurin_tag_17-jdk-amd64
ghcr.io/openhands/agent-server:bd0530f-java-arm64
ghcr.io/openhands/agent-server:bd0530f-eclipse-temurin_tag_17-jdk-arm64
ghcr.io/openhands/agent-server:bd0530f-python-amd64
ghcr.io/openhands/agent-server:bd0530f-nikolaik_s_python-nodejs_tag_python3.13-nodejs22-amd64
ghcr.io/openhands/agent-server:bd0530f-python-arm64
ghcr.io/openhands/agent-server:bd0530f-nikolaik_s_python-nodejs_tag_python3.13-nodejs22-arm64
ghcr.io/openhands/agent-server:bd0530f-golang
ghcr.io/openhands/agent-server:bd0530f-java
ghcr.io/openhands/agent-server:bd0530f-python

About Multi-Architecture Support

Each variant tag (e.g., bd0530f-python) is a multi-arch manifest supporting both amd64 and arm64
Docker automatically pulls the correct architecture for your platform
Individual architecture tags (e.g., bd0530f-python-amd64) are also available if needed

…tions When the agent posts messages, comments, issues, or other human-readable content to external services (Slack, GitHub, Discord, Linear, Notion, etc.), it should always include a brief note indicating the content was generated by an AI agent on behalf of the user. This applies universally — whether through dedicated tools, MCP integrations, or direct API calls — ensuring transparency regardless of the communication channel. Co-authored-by: openhands <openhands@all-hands.dev>

github-actions · 2026-03-17T06:11:54Z

Python API breakage checks — ✅ PASSED

Result: ✅ PASSED

Action log

github-actions · 2026-03-17T06:12:03Z

REST API breakage checks (OpenAPI) — ✅ PASSED

Result: ✅ PASSED

Action log

all-hands-bot

🟡 Taste Rating: Acceptable

This solves a real transparency problem with a simple, pragmatic approach. The instruction is comprehensive and covers the right cases.

⚠️ Eval Risk Flag

This PR modifies the system prompt template, which directly affects agent behavior. Per repo guidelines, prompt changes require human review with lightweight evals before approval. A maintainer should verify this doesn't negatively impact benchmark performance.

Assessment

✅ Solving a real problem: AI-generated content appearing as human-written is genuinely misleading
✅ Simple implementation: Just adds a rule to the prompt, no complexity
✅ Comprehensive coverage: Lists platforms, covers all integration types
⚠️ Minor verbosity: The bullet point is quite long (~5 lines) - see inline comment

Verdict: Good change, but needs human approval due to prompt modification policy.

all-hands-bot · 2026-03-17T06:35:53Z

openhands-sdk/openhands/sdk/agent/prompts/system_prompt.j2

 <EXTERNAL_SERVICES>
 * When interacting with external services like GitHub, GitLab, or Bitbucket, use their respective APIs instead of browser-based interactions whenever possible.
 * Only resort to browser-based interactions with these services if specifically requested by the user or if the required operation cannot be performed via API.
+* **AI disclosure**: When posting messages, comments, issues, or any content to external services that will be read by humans (e.g., Slack messages, GitHub/GitLab comments, PR/MR descriptions, Discord messages, Linear/Jira issues, Notion pages, emails, etc.), always include a brief note indicating the content was generated by an AI agent on behalf of the user. For example, you could add a line like: _"This [message/comment/issue/PR] was created by an AI assistant (OpenHands) on behalf of [user]."_ This applies to any communication channel — whether through dedicated tools, MCP integrations, or direct API calls.


🟡 Suggestion: This bullet point is quite long (~5 lines of text). Consider breaking it into sub-bullets for readability:

Suggested change

* **AI disclosure**: When posting messages, comments, issues, or any content to external services that will be read by humans (e.g., Slack messages, GitHub/GitLab comments, PR/MR descriptions, Discord messages, Linear/Jira issues, Notion pages, emails, etc.), always include a brief note indicating the content was generated by an AI agent on behalf of the user. For example, you could add a line like: _"This [message/comment/issue/PR] was created by an AI assistant (OpenHands) on behalf of [user]."_ This applies to any communication channel — whether through dedicated tools, MCP integrations, or direct API calls.

* **AI disclosure**: When posting messages, comments, issues, or any content to external services that will be read by humans, always include a brief note indicating the content was generated by an AI agent on behalf of the user.

- **Platforms**: Slack, GitHub/GitLab (comments, PR/MR descriptions), Discord, Linear/Jira, Notion, email, etc.

- **Example format**: _"This [message/comment/issue/PR] was created by an AI assistant (OpenHands) on behalf of [user]."_

- **Scope**: This applies to any communication channel — dedicated tools, MCP integrations, or direct API calls.

That said, the current version is clear and complete. This is a minor style preference, not a blocking issue.

VascoSch92

Makes sense.

Sometimes I have a hard time understanding who is writing what 😄

xingyaoww marked this pull request as ready for review March 17, 2026 06:34

all-hands-bot reviewed Mar 17, 2026

View reviewed changes

VascoSch92 approved these changes Mar 17, 2026

View reviewed changes

xingyaoww merged commit 58e7ff3 into main Mar 17, 2026
35 checks passed

xingyaoww deleted the add-ai-disclosure-policy branch March 17, 2026 12:06

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(prompt): add AI disclosure policy for external service communications#2476

feat(prompt): add AI disclosure policy for external service communications#2476
xingyaoww merged 1 commit intomainfrom
add-ai-disclosure-policy

xingyaoww commented Mar 17, 2026 •

edited by github-actions bot

Loading

Uh oh!

github-actions bot commented Mar 17, 2026

Uh oh!

github-actions bot commented Mar 17, 2026

Uh oh!

all-hands-bot left a comment

Uh oh!

all-hands-bot Mar 17, 2026

Uh oh!

VascoSch92 left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

-* **AI disclosure**: When posting messages, comments, issues, or any content to external services that will be read by humans (e.g., Slack messages, GitHub/GitLab comments, PR/MR descriptions, Discord messages, Linear/Jira issues, Notion pages, emails, etc.), always include a brief note indicating the content was generated by an AI agent on behalf of the user. For example, you could add a line like: _"This [message/comment/issue/PR] was created by an AI assistant (OpenHands) on behalf of [user]."_ This applies to any communication channel — whether through dedicated tools, MCP integrations, or direct API calls.
+* **AI disclosure**: When posting messages, comments, issues, or any content to external services that will be read by humans, always include a brief note indicating the content was generated by an AI agent on behalf of the user.
+  - **Platforms**: Slack, GitHub/GitLab (comments, PR/MR descriptions), Discord, Linear/Jira, Notion, email, etc.
+  - **Example format**: _"This [message/comment/issue/PR] was created by an AI assistant (OpenHands) on behalf of [user]."_
+  - **Scope**: This applies to any communication channel — dedicated tools, MCP integrations, or direct API calls.

Conversation

xingyaoww commented Mar 17, 2026 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Motivation

Why the SDK system prompt (not just individual skills)?

Changes

Impact

Uh oh!

github-actions bot commented Mar 17, 2026

Python API breakage checks — ✅ PASSED

Uh oh!

github-actions bot commented Mar 17, 2026

REST API breakage checks (OpenAPI) — ✅ PASSED

Uh oh!

all-hands-bot left a comment

Choose a reason for hiding this comment

Uh oh!

all-hands-bot Mar 17, 2026

Choose a reason for hiding this comment

Uh oh!

VascoSch92 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

xingyaoww commented Mar 17, 2026 •

edited by github-actions bot

Loading