Plugin-owned external verification approvals#15
Conversation
636a8b2 to
49f2895
Compare
|
Codex review: needs real behavior proof before merge. Reviewed June 18, 2026, 9:47 PM ET / 01:47 UTC. Summary Reproducibility: not applicable. this is a design RFC PR, not a bug report. The review checked the added RFC content, current main, GitHub discussion, and related RFC history instead of reproducing a runtime failure. Review metrics: 2 noteworthy metrics.
Merge readiness Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch. Rank-up moves:
Proof guidance:
Risk before merge
Maintainer options:
Next step before merge
Security Review detailsBest possible solution: Keep the RFC open as a draft until maintainers record the security/API decisions and discussion-thread outcome, add inspectable redacted proof where appropriate, then update the frontmatter and implementation tracking before merge. Do we have a high-confidence way to reproduce the issue? Not applicable; this is a design RFC PR, not a bug report. The review checked the added RFC content, current main, GitHub discussion, and related RFC history instead of reproducing a runtime failure. Is this the best way to solve the issue? Unclear until maintainer review; the command-template plus ownership-checked resolver is a coherent narrow design, but the RFC itself leaves security and compatibility decisions that should be settled before acceptance. AGENTS.md: not found in the target repository. Codex review notes: model internal, reasoning high; reviewed against e7555fdda445. Label changesLabel changes:
Label justifications:
Evidence reviewedSecurity concerns:
What I checked:
Likely related people:
What the crustacean ranks mean
Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics. How this review workflow works
|
kevinslin
left a comment
There was a problem hiding this comment.
rfcs should be high signal - please take a human pass through it and remove noise that doesn't need to be in here
| 6. AgentKit resolves its own pending approval. | ||
| 7. OpenClaw continues the original blocked tool call. | ||
|
|
||
| The first AgentKit attempt put World ID HITL directly in OpenClaw core: |
There was a problem hiding this comment.
we don't need this in the rfc - motivation should just be that certain plugins like worldcoin need ability to handle approvals
| approval flows. | ||
| - Letting auto review approve an external-proof-required approval without the | ||
| external proof. | ||
| - Designing durable chat status cards in this RFC. Those are useful but should |
There was a problem hiding this comment.
this is unrelated to plugin owning verification. we don't need it here. could we do a human pass and avoid slop in rfcs as much as possible?
| - it can classify whether the underlying command/tool call is risky | ||
| - it can explain why the action should be denied or escalated | ||
| - it can decline and fall back to the human operator | ||
| - it can include `externalVerificationRequired: true` in any bounded review |
There was a problem hiding this comment.
does this mean auto review can move an approval to the external verification lane? that would require a change to autoreview
| - channel-delivered approval prompts | ||
| - opt-in auto mode for lower-risk exec approval review | ||
|
|
||
| The safer-auto-mode blog frames the current direction well: |
There was a problem hiding this comment.
lets not reference a blog without any links. its also not necessary to call this out in the rfc since this is about plugin approval, not auto review
|
|
||
| ## Unresolved Questions | ||
|
|
||
| - Should verified resolution use the existing `operator.approvals` scope, or |
There was a problem hiding this comment.
we should answer these questions instead of leaving them open unless they are truly unresolvable rigth now
Summary
Adds a draft RFC for plugin-owned external verification of approval resolution. The proposal keeps OpenClaw core as the approval owner, avoids arbitrary plugin-defined approval actions, and introduces a narrow
externalResolutioncommand-template route plus an ownership-checked verified resolver.Motivation
This captures the AgentKit / World verification use case as a ClawHub plugin path instead of bundled core code, while addressing maintainer feedback that the earlier custom-actions shape expanded the API/UI/channel surface too much.
Validation
git diff --check --no-index -- /dev/null rfcs/0011-plugin-owned-external-verification-approvals.mdNotes
The RFC references the AgentKit plugin as the concrete reference implementation and calls out the interaction with auto approvals / auto review from the safer auto-mode design.