Skip to content

Track OpenClaw seams needed for full imported-plugin parity #1

Open

Description

@RomneyDa

What Problem This Tracks

Babelfish can translate many external plugin contributions using OpenClaw's existing public plugin SDK. Several remaining compatibility gaps cannot be implemented faithfully inside Babelfish because the relevant behavior is owned by OpenClaw, or because OpenClaw already has an internal bundle implementation that should be exposed rather than duplicated.

OpenClaw Public Seams Needed

  • Imported agent profiles: a public registration contract for named agent definitions with model selection, tool allow/deny policy, turn limits, memory, background mode, and workspace isolation. registerAgentHarness represents an entire runtime and is not an appropriate mapping for one imported agent definition.
  • Approval-boundary hooks: a public hook that fires when an interactive tool approval is about to be shown and can allow, deny, rewrite, or return a durable permission decision. Pre-tool policy can deny but cannot faithfully approve or update approval rules.
  • Bundle LSP projection: a public way for a native plugin to contribute bundle-derived LSP defaults. OpenClaw already supports Claude bundle LSP configuration internally; Babelfish should reuse that owner instead of launching a second LSP implementation.
  • Bundle settings projection: a public way to contribute validated bundle settings/defaults, including the supported Claude agent and subagentStatusLine settings. This should reuse OpenClaw's existing bundle settings merge behavior.
  • Connector registry: a public connector registration and authentication contract before Codex app connector IDs can be mapped. Connector IDs must not be treated as MCP endpoints.
  • Host notification events: a generic notification lifecycle event before Claude Notification hooks can be mapped without producing false events from unrelated message delivery.

Better Delegated To OpenClaw

  • Marketplace resolution and lifecycle: OpenClaw already owns bundle installation and Claude marketplace resolution, including source policy, caching, updates, enable/disable state, and managed restrictions. Babelfish should eventually delegate or forward to that infrastructure instead of implementing a second marketplace manager. Equivalent Codex or Hermes resolvers should also plug into the OpenClaw install owner when those ecosystems have stable marketplace contracts.

Why These Should Not Be Approximated In Babelfish

Approximations would either execute behavior at the wrong lifecycle boundary, bypass OpenClaw policy and authentication ownership, duplicate internal bundle loaders, or expose source capabilities with materially different security semantics. Babelfish should keep unsupported contributions visible until a stable host contract exists.

Desired Outcome

Add generic public SDK or manifest projection contracts for the capabilities above, then migrate Babelfish mappings to those contracts. The APIs should remain source-format-neutral so Codex, Claude Code, Hermes Agent, and future clients can share them.

Evidence

  • Babelfish support matrix documents the affected No and Partial surfaces.
  • OpenClaw already contains internal Claude bundle LSP/settings and marketplace behavior, but these are not dynamic public native-plugin registration surfaces.
  • OpenClaw pre-tool policy and agent harness APIs do not represent approval dialogs or individual imported agent profiles.

Activity

  1. added
    P3Low-risk cleanup, docs, polish, ergonomics, or speculative feature.
    impact:securityThis issue is about security boundaries, credentials, authz, sandboxing, or sensitive data.
    clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.
    clawsweeper:needs-security-reviewClawSweeper marked this issue as needing security-sensitive review.
    on Jul 13, 2026
  2. clawsweeper commented on Jul 13, 2026

    @clawsweeper

    Codex review: keeping this open for maintainer follow-up; there is still a little grit to resolve. Reviewed August 2, 2026, 3:11 PM ET / 19:11 UTC.

    Summary
    Keep open. Current main deliberately detects these imported-plugin surfaces, preserves them as unsupported, and warns at runtime; the requested source-neutral OpenClaw contracts remain unimplemented. This MEMBER-authored, security-sensitive umbrella needs a maintainer-sponsored host API and trust-boundary decision before Babelfish work can be safely scoped.

    Reproducibility: not applicable. this is a request for new host SDK contracts, not a report of broken established behavior. Current source and documentation consistently show that the named surfaces are intentionally detected and reported as unsupported.

    Ways to help us reproduce this

    • Add a screenshot or short recording showing the behavior.
    • Add expected vs actual behavior.

    Maintainer decision needed

    Question Recommendation
    Should OpenClaw sponsor a staged public SDK or manifest contract for these imported-plugin seams, and which seam should be the first narrowly scoped, policy-safe contract? Sponsor a staged host API: Choose one seam with clear ownership and define a source-neutral contract, authorization model, compatibility tests, and migration path before Babelfish maps it.

    Why: The issue combines new public API direction with approval, authentication, model, and execution trust boundaries; repository policy requires confirmation before manifest-contract changes, and Babelfish cannot safely choose those host semantics itself.

    Next step
    A maintainer must choose public host API and trust-boundary direction before the umbrella can be split into safe, implementable work.

    Review details

    Best possible solution:

    Keep the unsupported surfaces visible while OpenClaw sponsors narrow, source-neutral public contracts with explicit authorization and lifecycle semantics, then migrate each Babelfish mapping only after its host contract is approved and tested.

    Do we have a high-confidence way to reproduce the issue?

    Not applicable: this is a request for new host SDK contracts, not a report of broken established behavior. Current source and documentation consistently show that the named surfaces are intentionally detected and reported as unsupported.

    Is this the best way to solve the issue?

    Unclear: exposing a source-neutral contract is the right general direction, but the issue combines several lifecycle and security models that need separate host-owned API decisions before implementation.

    AGENTS.md: found and applied where relevant.

    Remaining risk / open question:

    • A generic catch-all seam could let imported plugins cross OpenClaw approval, authentication, model-selection, or execution-policy boundaries without a durable host-owned policy model.
    • The request bundles several independent contracts; landing them as one Babelfish-side abstraction would duplicate host ownership and make future compatibility guarantees harder to maintain.

    Codex review notes: model internal, reasoning high; reviewed against b5c42c46da2c.

    Label changes

    Label justifications:

    • P2: This is a meaningful compatibility and platform-design gap, but current behavior keeps the unsupported surfaces visible instead of causing an active outage.
    • impact:security: The requested approval, connector, authentication, model, and execution seams directly define plugin trust boundaries.
    Evidence reviewed

    What I checked:

    • Published compatibility boundary: The compatibility matrix still marks imported agent definitions as partial and approval hooks, notifications, Codex connectors, LSP servers, settings/default agents, and marketplace-native resolution as unsupported or unmapped. (README.md:95, 5e2ccd634980)
    • Current discovery preserves the gaps: Claude manifests are inspected for agents, LSP servers, and settings, but LSP/settings are recorded as unsupported; Codex app metadata is likewise recorded as an unsupported connector surface rather than converted into a different execution path. (src/bundle-plugins.ts:372, 5e2ccd634980)
    • Runtime explicitly warns instead of approximating: The runtime declares approval request/response hooks unsupported and logs every generated unsupported surface at startup, preserving the stated trust boundary rather than silently executing an approximation. (src/index.ts:103, 5e2ccd634980)
    • Release and current-main provenance: The baseline compatibility implementation is included in release v0.1.0, whose tag points to the release commit; current main is later at b5c42c46 and does not provide a released or main-only implementation of the requested host contracts. (CHANGELOG.md:10, b5c42c46da2c)
    • Repository policy supports retaining the boundary: The reviewed repository policy requires unsupported source behavior to remain visible, prohibits weakening trust-boundary warnings, and requires confirmation before public manifest-contract changes. (AGENTS.md:30, b5c42c46da2c)

    Likely related people:

    • steipete: Peter Steinberger authored the v0.1.0 release commit that introduced the compatibility matrix and the current unsupported-surface behavior. (role: introduced the released compatibility boundary; confidence: high; commits: 5e2ccd634980, 807a7c54e76f; files: README.md, src/index.ts, src/bundle-plugins.ts)
    • vincentkoc: Vincent Koc most recently changed the imported-hook runtime on current main, making them a relevant routing contact for any later hook-execution integration. (role: recent adjacent runtime contributor; confidence: medium; commits: b5c42c46da2c; files: src/index.ts)
    How this review workflow works
    • ClawSweeper keeps one durable marker-backed review comment per issue or PR.
    • Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
    • A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
    • PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
    • Maintainers can also comment @clawsweeper review to request a fresh review only.
    • Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
    • Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
    • Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.
  3. added
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    P2Normal priority bug or improvement with limited blast radius.
    impact:otherThis issue has meaningful maintainer-visible impact outside the owned taxonomy.
    clawsweeper:needs-security-reviewClawSweeper marked this issue as needing security-sensitive review.
    and removed
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    P3Low-risk cleanup, docs, polish, ergonomics, or speculative feature.
    clawsweeper:needs-security-reviewClawSweeper marked this issue as needing security-sensitive review.
    on Jul 29, 2026
  4. 1 remaining item

  5. added
    clawsweeper:needs-security-reviewClawSweeper marked this issue as needing security-sensitive review.
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    impact:otherThis issue has meaningful maintainer-visible impact outside the owned taxonomy.
    and removed
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    clawsweeper:needs-security-reviewClawSweeper marked this issue as needing security-sensitive review.
    impact:otherThis issue has meaningful maintainer-visible impact outside the owned taxonomy.
    on Jul 31, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    P2Normal priority bug or improvement with limited blast radius.clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.impact:securityThis issue is about security boundaries, credentials, authz, sandboxing, or sensitive data.issue-rating: 🌊 off-meta tidepoolIssue quality rating does not apply to this item.

    Type

    No type

    Fields

    Priority

    None yet

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions