Skip to content

[Feature] Multi-agent output provenance tracking #65823

@smonett

Description

@smonett

Problem

When multiple agents or models contribute to a synthesis (e.g., one model writes, another reviews, a third judges), the final output may claim multi-model validation. There is no native mechanism to verify that a "multi-model synthesis" actually involved multiple models — an agent could claim consensus without it existing.

Workaround

I built cc-synthesis-hook.py, a PreToolUse Write hook that guards writes to protected paths (governance docs, project files). It validates that content claiming multi-model input includes provenance markers referencing verified debate manifests (timestamped records of which models participated and what they produced).

Proposed Solution

Track multi-agent provenance natively:

  • When multiple agents/models contribute to output, record which models participated and what they produced
  • Expose provenance metadata in session transcripts
  • Optionally validate provenance claims before allowing writes to protected paths

Impact

Low. Relevant for advanced multi-agent workflows where output integrity matters. More of a research-grade feature than a daily-use need.

Environment

  • OpenClaw 2026.4.10 (npm, macOS)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions