1. src/agents/memory-search.ts, line 54-66 (link)

   Deprecation warning fires on every search / config resolution

   mergeConfig() is called each time a ResolvedMemorySearchConfig is needed (i.e. every search operation), so the deprecation log.warn(...) will fire repeatedly for anyone still using agents.defaults.memorySearch.extraPaths. There is no "warn once" guard, which will flood logs until the user migrates. Consider using a module-level flag or a Set to emit the warning only once per process:

   let extraPathsDeprecationWarned = false;
   
   // inside the if block:
   if (!extraPathsDeprecationWarned) {
     extraPathsDeprecationWarned = true;
     log.warn("agents.defaults.memorySearch.extraPaths is deprecated; ...");
   }

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: src/agents/memory-search.ts
   Line: 54-66
   
   Comment:
   **Deprecation warning fires on every search / config resolution**
   
   `mergeConfig()` is called each time a `ResolvedMemorySearchConfig` is needed (i.e. every search operation), so the deprecation `log.warn(...)` will fire repeatedly for anyone still using `agents.defaults.memorySearch.extraPaths`. There is no "warn once" guard, which will flood logs until the user migrates. Consider using a module-level flag or a `Set` to emit the warning only once per process:
   
   ```ts
   let extraPathsDeprecationWarned = false;
   
   // inside the if block:
   if (!extraPathsDeprecationWarned) {
     extraPathsDeprecationWarned = true;
     log.warn("agents.defaults.memorySearch.extraPaths is deprecated; ...");
   }
   ```
   
   How can I resolve this? If you propose a fix, please make it concise.

• 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.

• 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
• 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
• 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
• 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
• 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
• 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
• 🌊 off-meta tidepool: rating does not apply to this item.

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.

Best possible solution:

Rebuild the feature on the current memory-core plugin and host SDK boundary, preserving extraPaths compatibility unless a maintainer approves a tested migration, with strict shared-root validation, docs, and fresh plus upgrade CLI proof.

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

Yes for the review blockers by source inspection: current main uses the memory-core plugin/host SDK boundary, while the PR head still edits stale memory paths and contains the extraPaths and blank-path logic described above. I did not run the CLI because this review is read-only.

Is this the best way to solve the issue?

No. The submitted implementation is not the best path because it mixes a new cross-agent memory feature with stale core files and compatibility-sensitive config migration; the safer direction is a current plugin-boundary design with explicit validation, docs, and upgrade proof.

Label justifications:

• P2: This is a meaningful memory feature with limited blast radius, but it is not an urgent runtime regression.
• merge-risk: 🚨 compatibility: The diff changes existing default extraPaths indexing behavior and introduces a new reserved _shared agent ID/config surface.
• merge-risk: 🚨 session-state: The proposed store crosses agent memory boundaries and can alter which documents are visible to each agent.
• merge-risk: 🚨 security-boundary: The diff adds a new operator-configured file-read surface and currently accepts blank shared paths too loosely.
• rating: 🧂 unranked krab: Current PR rating is 🧂 unranked krab because proof is 🧂 unranked krab, patch quality is 🧂 unranked krab, and Not quality-ready yet: real behavior proof is missing and the patch has blocking compatibility, architecture, and security-boundary issues.
• status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The PR body lists manual checks but provides no inspectable terminal output, logs, screenshots, recordings, or linked artifacts; add redacted CLI proof and update the PR body so ClawSweeper can re-review.

Full review comments:

• [P1] Rebase shared store into the active memory plugin — src/memory/manager.ts:271-274
  Current main now runs memory CLI/runtime through extensions/memory-core and packages/memory-host-sdk, but this branch adds the _shared manager to the old src/memory implementation. After resolving conflicts, the shared index/search/status behavior needs to live on the active plugin/host SDK path or it will either be dead code or bypass the current architecture.
  Confidence: 0.9
• [P1] Preserve default extraPaths indexing semantics — src/agents/memory-search.ts:257-264
  This drops default extraPaths from the resolved local extraPaths and migrates them into shared memory. Current main explicitly maps default and per-agent memorySearch.extraPaths into agent-scoped QMD collections, so existing configs can silently change indexing topology and collection names on upgrade.
  Confidence: 0.91
• [P1] Reject blank sharedPaths before read authorization — src/config/zod-schema.agent-runtime.ts:596-602
sharedPaths accepts blank strings and the merge code resolves every entry before filtering. A blank path can reach the shared read check, where absPath.startsWith(${sp.path}${path.sep}) effectively matches absolute Markdown paths, widening the read boundary from a config typo.
  Confidence: 0.88
• [P2] Do not mkdir configured shared files — src/memory/manager.ts:248-252
  The constructor calls ensureDir for every configured shared path, but memory extra paths support both directories and individual .md files. A configured not-yet-created file such as policy.md would become a directory and block the intended file from being created/indexed.
  Confidence: 0.82
• [P2] Read shared status from the manager's real store path — src/memory/manager.ts:884-893
  Shared status reads settings.sharedStorePath, which is always ~/.openclaw/memory/_shared.sqlite, while the _shared manager itself uses the configured memorySearch.store.path with {agentId} support. Custom store paths can therefore report the wrong DB and stale zero counts.
  Confidence: 0.84

Overall correctness: patch is incorrect
Overall confidence: 0.89

Security concerns:

• [high] Blank shared paths can widen Markdown read access — src/config/zod-schema.agent-runtime.ts:596
  The schema accepts blank sharedPaths and the manager later uses direct path-prefix checks; a blank shared path can behave like a root prefix for absolute Markdown reads.
  Confidence: 0.88
• [medium] Cross-agent memory requires explicit boundary review — src/memory/manager.ts:730
  The _shared store lets agents search/read files outside their own workspace when configured, so maintainers need docs, validation, and proof that the operator-controlled boundary cannot be broadened by malformed config.
  Confidence: 0.82

What I checked:

• Current checkout state: The target checkout was clean on current main before review. (5c4a733912a8)
• Live PR state: GitHub reports the PR as DIRTY/CONFLICTING and the real-behavior-proof checks are failing, so it cannot merge as-is even before code review findings. (35b27aa65a1f)
• Current memory CLI location: Current main registers memory CLI commands in the memory-core plugin; there is no --shared path in the active status or index commands here. (extensions/memory-core/src/cli.ts:116, 5c4a733912a8)
• Current active memory manager boundary: Current main's memory manager lives under extensions/memory-core and delegates file reads through the host SDK, so the PR's src/memory/manager.ts changes are on the old core implementation boundary. (extensions/memory-core/src/memory/manager.ts:1, 5c4a733912a8)
• Current extraPaths contract: Current main merges default and per-agent memorySearch.extraPaths into QMD collections, and adjacent tests assert those defaults remain agent-scoped collections. (packages/memory-host-sdk/src/host/backend-config.ts:399, 5c4a733912a8)
• PR changes extraPaths semantics: The PR migrates default extraPaths into shared paths and then builds local extraPaths only from overrides, which would silently change existing default extra-path indexing behavior. (src/agents/memory-search.ts:234, 35b27aa65a1f)

Likely related people:

• steipete: Peter Steinberger moved the memory engine into the memory-core plugin and host SDK package, and has many recent touches across the active memory manager/config files. (role: recent architecture owner; confidence: high; commits: cad83db8b2f7, eebce9e9c7cb, 77e6e4cf87; files: extensions/memory-core/src/memory/manager.ts, packages/memory-host-sdk/src/host/backend-config.ts, src/agents/memory-search.ts)
• vincentkoc: Vincent Koc merged and authored recent QMD collection and memory-core fixes, and is the current assignee on this PR. (role: recent area contributor and merger; confidence: high; commits: ab4ddff7f1e2, b7de04f23f, da35718cb2; files: packages/memory-host-sdk/src/host/backend-config.ts, extensions/memory-core/src/memory/manager.ts)
• Vitalcheffe: Authored the merged PR that wired memorySearch.extraPaths into QMD indexing, which is the compatibility contract this PR would change. (role: introduced adjacent extraPaths behavior; confidence: medium; commits: 219d4f03bd57; files: packages/memory-host-sdk/src/host/backend-config.ts)

Codex review notes: model gpt-5.5, reasoning high; reviewed against 5c4a733912a8.

• The egg game starts only after the PR passes the real-behavior proof check.
• Before that, no creature or rarity is rolled. The treat waits for real proof.
• This is still just collectible flavor: proof affects review readiness, not creature quality.