Stock prompt plugin for plugin-spec prompt automations

Open
#800 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

  • #799 by @simonrosenberg — closed without merging

Assessment

Difficulty
3/5
Estimated time
1-2 days
Newbie friendliness
54/100
Issue type
Feature
Clarity
Clearly specified
Activity status
Active
Tech stack
python

Research direction

Start by reading the existing plugin manifests and conversation-runner commands under plugins/, then inspect the plugin contract tests. Add plugins/prompt-task with a required prompt config and required agent profile, and cover the manifest and run-message contract in a test. Done means the configured prompt is passed as the first run message, the chosen profile is required, and the listed repository checks pass.

Written by the indexing model from the issue text.

Description

enhancement needs-automation-release priority:low ready-for-dev
Problem

"Create Automation → enter a prompt" runs ignore the user's agent profile, so a profile with sub-agents runs without the task tool (OpenHands/automation#557). Prompt automations are created through the prompt preset, which never reads a profile, and presets are frozen in favor of plugin-spec automations (OpenHands/automation#600).

Plugin-spec automations do launch from an agent profile, but there is no stock plugin that carries a per-automation prompt, so the prompt flow can't move onto them.

Proposed change

Add plugins/prompt-task, a conversation-runner plugin:

  • Its command takes the automation's prompt as arguments: entrypoint.arguments: "{{ config.prompt }}", rendered by the Automation Service (OpenHands/automation#615).
  • runtime.agent_profile.required: true with no default_name, so the creator picks the profile.
  • config_schema takes one required prompt.
Acceptance criteria
  • A plugin-spec automation created from plugins/prompt-task with a prompt and an agent_profile_id runs the prompt on a conversation launched from that profile.
  • The run's first message (/prompt-task:run <prompt>) triggers no public skill.
  • Repo checks pass (pytest tests/, scripts/sync_extensions.py --check, scripts/check_plugin_resources.py).

Implementation: #799.


I'm an AI agent (Claude Code) helping @simonrosenberg.


OpenHands AI triage

The following comments and acceptance criteria were added by the OpenHands AI agent.

Triage

Add plugins/prompt-task to OpenHands/extensions: a conversation-runner automation plugin whose run message carries a per-automation prompt and whose run launches from the agent profile the creator picks. The extensions-owned deliverable is the plugin (manifest, command, README, vendor manifests) and a contract test; the Automation Service pieces it depends on are OpenHands/automation#614 (dispatch launches the run's conversation from the profile) and OpenHands/automation#615 (agent_profile_id on POST /v1/plugin-spec plus {{ config.* }} argument rendering), and Canvas creation is OpenHands/OpenHands. Scope is bounded to the new plugin directory and its test.

Non-goals: no changes to the prompt or plugin presets; no Automation Service, SDK, or Canvas code; no marketplace/catalog listing (this plugin is prompt-flow infrastructure, so the sync_extensions.py coverage warning for it is expected and non-blocking); no tool, secret, or MCP allowlists on the profile requirement, since the plugin inherits the selected profile's own settings.

Acceptance Criteria
  • plugins/prompt-task exists as an executable automation plugin: .plugin/plugin.json validates against the Automation Service manifest schema (openhands.automation.plugin.v1) with entrypoint.runner: "conversation", an entrypoint.command whose commands/<command>.md file exists, and entrypoint.arguments that renders config.prompt rather than emitting the literal {{ config.prompt }}.
  • runtime.agent_profile is required: true with no default_name, so creating a plugin-spec automation from this plugin with a prompt and an agent_profile_id is accepted, and creating one without agent_profile_id is rejected.
  • config_schema requires exactly one prompt (non-empty string, documented maximum length) and rejects a config that omits prompt.
  • A dispatched run's first user message is /<plugin-name>:<command> <prompt>, containing the configured prompt text; the run's conversation reports launched_agent_profile as the chosen profile, and the tools offered to the model include that profile's (for example task when the profile enables sub-agents).
  • The run message triggers no public skill: no triggers: entry in any skills/*/SKILL.md, plugins/*/SKILL.md, or plugins/*/skills/*/SKILL.md matches the run message (case-insensitive).
  • .claude-plugin and .codex-plugin resolve to .plugin (symlink or byte-identical plugin.json mirror).
  • Repo checks pass: pytest tests/, python scripts/sync_extensions.py --check, python scripts/check_plugin_resources.py.
  • The README states the Automation Service version requirement (pluginSpecAgentProfiles in GET /v1/capabilities) and that an older service rejects the plugin at create instead of running a literal {{ config.prompt }}.
Dominant language
Python
Stars
184
Forks
137
Avg merge
1d 19h
Merged PRs (30d)
46

Getting set up

  • No Dockerfile or Docker Compose file
  • Has a pull request template
  • No contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from OpenHands/extensions

All issues in OpenHands/extensions

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.