Summary

• Problem: image/video provider inventory tool output can be visible on shared chat surfaces.
• Why it matters: provider availability, auth hints, and operator-environment details belong in direct diagnostics, not group/channel delivery.
• What changed: tag image/video provider inventory tool results, suppress that tagged text on non-direct surfaces before tool TTS runs, and preserve generated media plus direct-chat diagnostics.
• What did NOT change (scope boundary): normal tool summaries, direct-chat verbose output, media delivery, provider selection, and generation behavior are unchanged.

Change Type (select all)

• Bug fix
• Feature
• Refactor required for the fix
• Docs
• Security hardening
• Chore/infra

Scope (select all touched areas)

• Gateway / orchestration
• Skills / tool execution
• Auth / tokens
• Memory / storage
• Integrations
• API / contracts
• UI / DX
• CI/CD / infra

Linked Issue/PR

• Closes [Bug]: image/video provider inventory leaks into Discord group channels as visible tool output #75166
• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: completed image/video provider inventory results were treated like ordinary visible tool output after agent execution.
• Missing detection / guardrail: the dispatcher had no typed way to distinguish internal provider inventory from normal tool result text before shared-surface delivery and tool TTS synthesis.
• Contributing context (if known): verbose/full tool-output delivery is useful in direct diagnostics, but shared chat delivery needs a stricter rule for internal operator-environment inventory.

Regression Test Plan (if applicable)

• Coverage level that should have caught this:
  • Unit test
  • Seam / integration test
  • End-to-end test
  • Existing coverage already sufficient
• Target test or file: src/agents/tool-output-shapes.test.ts, src/auto-reply/reply/dispatch-from-config.test.ts, src/agents/pi-embedded-subscribe.handlers.tools.media.test.ts
• Scenario the test should lock in: image/video provider inventory is tagged, non-direct delivery suppresses the text before tool TTS, generated media still delivers, and direct chats still receive diagnostics.
• Why this is the smallest reliable guardrail: the classifier test locks the producer shape, and the dispatcher tests lock the shared-surface delivery seam where text/TTS/media decisions happen.
• Existing test that already covers this (if any): media handler coverage validates the existing image/video result path still works.
• If no new test is added, why not: N/A

User-visible / Behavior Changes

Shared chat surfaces no longer receive image/video provider inventory text. Direct chats keep that diagnostic output visible.

Diagram (if applicable)

Before:
image/video provider inventory -> ordinary tool result -> shared text/TTS delivery

After:
image/video provider inventory -> tagged internal shape -> shared text/TTS suppressed, media preserved

Security Impact (required)

• New permissions/capabilities? (No)
• Secrets/tokens handling changed? (No)
• New/changed network calls? (No)
• Command/tool execution surface changed? (No)
• Data access scope changed? (No)
• If any Yes, explain risk + mitigation: N/A

Repro + Verification

Environment

• OS: Linux
• Runtime/container: Node 22 with repo pnpm
• Model/provider: mocked image/video tool result payloads
• Integration/channel (if any): Discord channel and Telegram group/direct dispatcher contexts
• Relevant config (redacted): automatic group reply config with TTS mode exercised in tests

Steps

1. Emit an image_generate or video_generate list result with details.providers.
2. Deliver it through a non-direct chat with tool TTS enabled.
3. Deliver the same tagged payload through a direct chat.

Expected

• Non-direct chats suppress provider inventory text before TTS.
• Non-direct media-only payloads still deliver media.
• Direct chats keep provider inventory diagnostics visible.

Actual

• Matches expected with the new dispatcher and classifier coverage.

Evidence

• Trace/log snippets

Commands run locally:

• pnpm test src/agents/tool-output-shapes.test.ts src/auto-reply/reply/dispatch-from-config.test.ts --run
• pnpm test src/agents/pi-embedded-subscribe.handlers.tools.media.test.ts --run
• pnpm changed:lanes --base refs/tmp/openclaw-main-current --json
• pnpm check:changed --base refs/tmp/openclaw-main-current
• pnpm exec oxfmt --check --threads=1 CHANGELOG.md docs/tools/thinking.md src/agents/pi-embedded-subscribe.ts src/agents/runtime-plan/types.ts src/agents/tool-output-shapes.ts src/agents/tool-output-shapes.test.ts src/auto-reply/reply-payload.ts src/auto-reply/reply/dispatch-from-config.ts src/auto-reply/reply/dispatch-from-config.test.ts
• git diff --check refs/tmp/openclaw-main-current...

Human Verification (required)

What you personally verified (not just CI), and how:

• Verified scenarios: provider inventory classifier, non-direct pre-TTS suppression, non-direct media preservation, direct-chat diagnostics, existing image/video media handler behavior, changed-lane gate.
• Edge cases checked: unrelated tools with details.providers are not tagged; image/video media results without provider inventory are not tagged; group/channel suppression strips text and spokenText while preserving media.
• What you did not verify: live Discord/Telegram delivery against real accounts.

Review Conversations

• I replied to or resolved every bot review conversation I addressed in this PR.
• I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

• Backward compatible? (Yes)
• Config/env changes? (No)
• Migration needed? (No)
• If yes, exact upgrade steps: N/A

Risks and Mitigations

• Risk: shared chats may lose a diagnostic provider inventory message that was useful during debugging.
  • Mitigation: direct chats keep the diagnostic visible, and generated media still delivers on shared surfaces.