feat: allow tool.execute.after hooks to inject AI-visible messages

[doomsday616]

Issue for this PR

Closes#17412

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

Adds an optional inject field to the tool.execute.after hook output. When a plugin sets output.inject, the messages are persisted as synthetic user messages that the AI sees on the next loop iteration.

Why this matters — and why we wrote it ourselves:

We filed #17412 two weeks ago, tagged @thdxr, @adamdotdevin, @nexxeln, @Hona, @jayair, @fwang, and @kujtimiihoxha. Zero response. The broader #12472 (Claude Code hooks compat) has 10 👍, 11 comments from the community, and also zero team response. So here we are — writing it ourselves.

This isn't a nice-to-have. Without hook message injection, any skill requiring behavioral enforcement is non-viable on OpenCode. The planning-with-files benchmark tells the story:

Metric	With hooks (Claude Code)	Without hooks
Pass rate (30 assertions)	96.7% (29/30)	6.7% (2/30)
3-file planning pattern followed	5/5	0/5

That's a 15x degradation. It means OpenCode cannot reliably handle complex, multi-step projects that need structured planning — the AI forgets its workflow after ~5 tool calls and completely loses it after compaction. Users in #12472 have explicitly said they recommend Cline over OpenCode for serious work because of this gap.

If this PR also gets ignored, frankly, it's hard to see how OpenCode's skill ecosystem will ever be competitive.

The actual change:

A flushInjectedMessages() helper creates synthetic user messages using the same Session.updateMessage() + Session.updatePart() pattern already used by subtask summary messages in prompt.ts. It's called after all three tool.execute.after hook sites (registry tools, MCP tools, subtask tools). System-role injections get wrapped in <system-reminder> tags.

~57 lines across 2 files. Fully backward compatible — inject is optional.

Usage:

"tool.execute.after": async(input,output)=>{if(input.tool==="edit"){output.inject=[{role: "user",text: "Remember to update progress.md."},]}}

How did you verify your code works?

The implementation reuses the exact synthetic user message pattern from prompt.ts (lines 524-542, subtask summary injection) — same Session.updateMessage() + Session.updatePart() path, same synthetic: true flag. TypeCheck passes. Relying on CI for full validation as this was authored remotely via GitHub API.

Screenshots / recordings

N/A — no UI changes.

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍

[rekram1-node]

Automated PR Cleanup

Thank you for contributing to opencode.

Due to the high volume of PRs from users and AI agents, we periodically close older PRs using automated criteria so maintainers can focus review time on the most active and community-supported contributions.

This PR was closed because it matched the following cleanup criteria:

• The PR was created more than 1 month ago
• The PR had fewer than 2 positive reactions
• Positive reactions are counted as thumbs-up, heart, celebration, or rocket reactions on the PR

PRs created within the last month are not affected by this cleanup.

If you believe this PR was closed incorrectly, or if you are still actively working on it, please leave a comment explaining why it should be reopened. A maintainer can review and reopen it if appropriate.

Thanks again for taking the time to contribute.