• 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:

Land the targeted Codex/memory fix after a maintainer explicitly accepts the read/write-only memory-flush contract and required checks pass.

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

Yes, from source inspection rather than an executed repro. Current main forwards memoryFlushWritePath into tool construction but still lets the Codex app-server path reach native/sandbox shell tooling while the shared Pi contract restricts memory runs to read plus append-only write.

Is this the best way to solve the issue?

Yes. The branch fixes both source-backed causes by pre-creating the canonical file before the maintenance run and aligning Codex dynamic/native/sandbox tool selection with the existing memory-flush contract, subject to maintainer acceptance of the compatibility change.

Label justifications:

• P1: The PR addresses an observed Codex memory flush failure during compaction that surfaced through a real channel workflow.
• merge-risk: 🚨 compatibility: The diff deliberately removes native shell/file and sandbox shell tools from Codex memory-triggered runs, which can break existing custom prompts that relied on that surface.
• rating: 🐚 platinum hermit: Current PR rating is 🐚 platinum hermit because proof is 🌊 off-meta tidepool, patch quality is 🐚 platinum hermit, and Good normal bug-fix PR with focused tests and no blocking findings; merge readiness depends on maintainer acceptance of the compatibility change.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Not applicable: The external contributor proof gate does not apply because this is a MEMBER PR with the protected maintainer label; the PR body lists targeted regression, type, lint, and diff checks.

What I checked:

• Current main Codex tool surface: Current main builds Codex dynamic tools, filters app-server-owned tools, then can re-add sandbox_exec/sandbox_process whenever sandboxing is active and the native surface is disabled; the current code has no memory-trigger special case. (extensions/codex/src/app-server/run-attempt.ts:3603, 03125c8e132d)
• Shared memory flush contract: The shared OpenClaw tool factory already treats trigger=memory plus memoryFlushWritePath as a constrained run, requiring the write path and exposing only read plus append-only write before later policy filters. (src/agents/pi-tools.ts:475, 03125c8e132d)
• Canonical memory target: The memory-core flush plan directs writes to the daily canonical memory/YYYY-MM-DD.md path and tells the agent to create memory/ if needed and append rather than overwrite. (extensions/memory-core/src/flush-plan.ts:13, 03125c8e132d)
• PR implementation: The PR adds ensureMemoryFlushTargetFile before the maintenance run, disables Codex native tool surface for memory flushes, filters memory flush dynamic tools to read/write, and blocks sandbox_exec from being re-added for memory flushes. (extensions/codex/src/app-server/run-attempt.ts:3581, c4d747f6d816)
• Target pre-create implementation: The PR validates the memory flush relative path lexically under the workspace, creates the parent directory, opens the target with append mode, and calls that before runWithModelFallback starts the flush turn. (src/auto-reply/reply/agent-runner-memory.ts:89, c4d747f6d816)
• Regression coverage: The PR adds tests that assert the target file is ensured before the embedded agent runs and that Codex memory flush dynamic tools resolve to only read/write with native surface disabled. (extensions/codex/src/app-server/run-attempt.test.ts:763, c4d747f6d816)

Likely related people:

• @TurboTheTurtle: Blame shows c2004fe owns the current memory flush runner handoff and much of the Codex app-server dynamic tool build path that this PR adjusts. (role: recent area contributor; confidence: high; commits: c2004fe6622d; files: src/auto-reply/reply/agent-runner-memory.ts, extensions/codex/src/app-server/run-attempt.ts)
• @joshavant: Commit ba06376 recently changed the Codex native/sandbox tool-surface checks that control whether sandbox_exec is exposed. (role: recent sandbox/tool-surface contributor; confidence: high; commits: ba06376c7955; files: extensions/codex/src/app-server/run-attempt.ts)
• @frankekn: History shows prior memory-flush contract work forwarding memoryFlushWritePath and protecting bootstrap files during memory flush runs. (role: memory flush contract contributor; confidence: medium; commits: 8306eabf85ea, 96e4975922de; files: src/agents/pi-tools.ts, src/auto-reply/reply/agent-runner-memory.ts, src/agents/pi-embedded-runner/run.ts)
• @amknight: The author also has recent merged current-main work in the same Codex app-server run-attempt surface, beyond this proposed branch. (role: recent adjacent contributor; confidence: medium; commits: cc4e30b3d9c2; files: extensions/codex/src/app-server/run-attempt.ts)

Codex review notes: model gpt-5.5, reasoning high; reviewed against 03125c8e132d.

• Eggs appear after the PR passes real-behavior proof. It is here for vibes, not verdicts: it does not change labels, ratings, merge decisions, or automation.
• The shell reacts to review momentum: open follow-up work warms it up, re-review makes it wobble, and a clean final review lets it hatch.
• Hatchability usually comes from sufficient real-behavior proof, no blocking P0/P1/P2 findings, no security attention needed, and clean correctness. A merged PR is already final, so merge makes the egg hatchable independently.
• The hatch is seeded from this repository and PR number, so the same PR keeps the same creature; the reviewed head SHA can only change safe visual details.
• Rarity is just collectible sparkle: 🥚 common, 🌱 uncommon, 💎 rare, ✨ glimmer, and 🌈 legendary.