Fix Claude never-prompt Keychain cache bypass (#1991)

[Yuxin-Qiao]

Summary

• When Claude prompt mode is stored as never (Avoid Keychain prompts), CodexBar no longer reads/writes/clears the CodexBar-owned oauth.claude Keychain cache during OAuth load, invalidation, or cache presence checks.
• Background refresh under never now prefers ~/.claude/.credentials.json (and existing Claude-keychain paths governed by the never policy) instead of repeatedly touching the CodexBar OAuth cache item that can surface macOS prompts.
• CodexBar-owned memory cache entries are also bypassed under never, so a prior prompt-enabled load cannot keep serving stale oauth.claude cache data after the user switches modes.
• Deferred invalidation uses a pending-clear tombstone: explicit logout / file-change invalidation under never records the pending marker and flushes it once prompt-enabled cache access is safe again (retrying on locked-keychain failures).
• MCP-only detection stays available for the experimental security(1) reader (effectiveMode(readStrategy:)), while Security.framework probes remain blocked under stored never.
• Owner-resolution probes of Claude Code-credentials are skipped when stored prompt mode is never.
• Adds focused regression tests in ClaudeOAuthCredentialsStoreNeverPromptCacheTests plus updates to ownership / security-CLI suites.

Fixes the bug reported in #1991 where periodic refresh could still trigger macOS Keychain UI even after choosing Avoid Keychain prompts. #1990 (manual-refresh pre-alert UX) is intentionally unchanged.

Codex review follow-ups

Review item	Resolution
Prevent stale cache reuse after never-mode invalidation	clearCacheKeychain() records pendingCodexBarOAuthKeychainCacheClear under never; flush on re-enable (28f7c02b, tests)
Ignore cache-derived memory entries in never mode	shouldUseCodexBarOAuthKeychainCache gates memory + cache reads (3c340b40, tests)
Update stale never-mode cache ownership test	load record ignores codexbar cache in never prompt mode; prior assertion moved to .onlyOnUserAction (e5ca4498)
Avoid reporting the same file change forever	shouldSaveFileFingerprint under never + idempotent second invalidateCacheIfCredentialsFileChanged() (3c340b40, tests)
Keep pending clear until keychain delete succeeds	flushPendingCodexBarOAuthKeychainCacheClearIfNeeded() retains tombstone on .failed; load returns .temporarilyUnavailable (e5ca4498, tests)
Preserve MCP-only detection for experimental reader	effectiveMode(readStrategy:) in isMcpOAuthOnlyClaudeKeychainPayloadPresent (e5ca4498, tests)
Keep pending clear until replacement store succeeds	saveToCacheKeychain clears tombstone only after successful storeResult (28f7c02b, tests)

Behavior proof (redacted, no live Keychain UI)

Repo policy (AGENTS.md) avoids live macOS Keychain prompt validation in CI/agents. Proof below is from isolated test-keychain runs on macOS 15 using per-test KeychainCacheStore.withServiceOverrideForTesting (no user Keychain prompts).

$ swift test --filter 'ClaudeOAuthCredentialsStoreNeverPromptCacheTests|ClaudeOAuthCredentialsStoreCLIStorageOwnershipTests|ClaudeOAuthCredentialsStoreIsolatedSecurityCLITests'
...
✔ Test "never prompt mode load prefers credentials file over codexbar oauth keychain cache in background" passed
✔ Test "never prompt mode load does not write codexbar oauth keychain cache" passed
✔ Test "never prompt mode invalidate credentials file changed leaves codexbar oauth keychain cache untouched" passed
✔ Test "never prompt mode has cached credentials ignores codexbar oauth keychain cache" passed
✔ Test "leaving never prompt mode clears stale codexbar oauth keychain cache after credentials file change" passed
✔ Test "never prompt mode ignores memory cache populated under prompt mode" passed
✔ Test "never prompt mode invalidate cache records pending clear for explicit logout" passed
✔ Test "pending oauth keychain cache clear retries when keychain delete temporarily unavailable" passed
✔ Test "save to cache keychain keeps pending clear when replacement store fails" passed
✔ Test "never prompt mode still detects MCP-only payload via experimental security CLI reader" passed
✔ Test "load record ignores codexbar cache in never prompt mode" passed
✔ Test run with 21 tests in 3 suites passed

Representative redacted logs from the never-prompt suite:

info com.steipete.codexbar.claude-usage: Claude OAuth credentials file changed; invalidating cache
info com.steipete.codexbar.keychain-cache: Keychain cache delete temporarily unavailable (oauth.claude)

The second line is from the simulated errSecInteractionNotAllowed retry path; under never the production oauth.claude item is not consulted until the pending clear succeeds.

Live macOS repro (installed app 0.41.0)

On a real setup against the installed 0.41.0 (100) app (not just isolated tests):

1. Seed oauth.claude under com.steipete.codexbar.cache
2. Set claudeOAuthKeychainPromptMode = never
3. Run CodexBarCLI usage --provider claude --source oauth

Observed: CLI hung for >=12s waiting for macOS Keychain UI while never mode was active.

Helper script (added in 4a41e582):

./Scripts/repro_keychain_issues.sh 2025

Details: #2011 (comment)

Test plan

• swift test --filter 'ClaudeOAuthCredentialsStoreNeverPromptCacheTests|ClaudeOAuthCredentialsStoreTests|ClaudeOAuthCredentialsStorePromptPolicyTests|ClaudeOAuthCredentialsStoreCLIStorageOwnershipTests|ClaudeOAuthCredentialsStoreIsolatedSecurityCLITests'
• make check (lint green on latest CI rollup)
• Redacted macOS behavior proof via isolated test-keychain + structured logs (above)
• Live macOS repro on installed 0.41.0 app + Scripts/repro_keychain_issues.sh

Maintainer-requested live proof is in the PR comment above; isolated tests remain the default CI path per AGENTS.md.

[clawsweeper]

Codex review: needs real behavior proof before merge. Reviewed July 9, 2026, 2:11 PM ET / 18:11 UTC.

Summary
The PR gates Claude OAuth CodexBar-owned Keychain cache reads, writes, clears, and memory-cache reuse when prompt mode is .never, adds pending-clear handling, preserves experimental security-CLI MCP detection, and adds focused Swift regression tests.

Reproducibility: yes. source-backed: the linked issue includes affected-machine diagnostics, and current main reads the CodexBar-owned Claude OAuth cache before file fallback even when prompt mode is .never. I did not run live Keychain validation because AGENTS.md forbids unrequested prompt-producing checks.

Review metrics: 2 noteworthy metrics.

• Changed surface: 2 source files, 3 test files. The implementation is concentrated in Claude OAuth cache and security-CLI behavior with matching provider-focused tests.
• Regression suite added: 586-line test file added. Most of the diff is focused coverage for never-mode cache reads, writes, invalidation, memory bypass, and pending-clear failure paths.

Root-cause cluster
Relationship: fixed_by_candidate
Canonical: #1991
Summary: This PR is the candidate fix for the canonical v0.41.0 periodic Claude Keychain prompt loop; the manual-refresh prompt report overlaps but is intentionally outside this PR.

Members:

• canonical: v0.41.0: macOS Keychain password prompt repeats every few minutes #1991 - The PR body says it fixes the periodic v0.41.0 Keychain prompt loop, and the issue discussion identifies the same .never plus oauth.claude cache path.
• partial_overlap: 每次手动刷新都会有提示框，这是为什么？ #1990 - The manual-refresh pre-alert report is related Claude Keychain UX, but the PR body explicitly says that manual-refresh pre-alert behavior is unchanged.
• duplicate: Claude logging out and persistent macOS Keychain prompt loop on v0.41.0 #1998 - The closed report describes the same v0.41.0 periodic Claude Keychain prompt loop, and steipete consolidated it into the canonical linked issue.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🦪 silver shellfish
Proof: 🦪 silver shellfish
Patch quality: 🦞 diamond lobster
Result: blocked until real behavior proof from a real setup is added.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

• [P1] Add redacted affected-app runtime proof, preferably unified logs across a refresh window showing no CodexBar oauth.claude Keychain activity while .never is set.
• Redact private data in any proof, including tokens, account identifiers, machine details, endpoints, and non-public paths.

Proof guidance:

• [P1] Needs real behavior proof before merge: The PR body includes copied focused Swift test output and simulated logs from isolated test-store/test-keychain runs, but no after-fix runtime proof from an affected CodexBar app setup. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

• [P1] The only after-fix proof in the PR body is copied Swift test output and simulated logs from isolated test-store/test-keychain runs, not redacted runtime evidence from an affected CodexBar app setup.
• [P1] The diff intentionally changes Claude OAuth cache behavior for existing .never users; maintainers should confirm the upgrade behavior for cache-only setups with no current credentials file is acceptable before merge.

Maintainer options:

1. Require affected-app proof (recommended)
   Before merge, require redacted runtime proof from an affected setup or maintainer machine showing .never no longer touches the CodexBar Claude OAuth cache during refresh.
2. Accept isolated coverage intentionally
   Maintainers may choose to accept the isolated test-store proof because repository policy avoids live Keychain prompt validation.
3. Hold for auth-policy review
   Pause merge if the cache-only upgrade behavior under .never needs a product or auth-policy decision first.

Next step before merge

• [P1] No automated repair lane is appropriate because the remaining blocker is contributor or maintainer proof sufficiency for an auth-sensitive Keychain cache change.

Maintainer decision needed

• Question: For this auth-sensitive Claude Keychain cache fix, should maintainers require affected-app runtime proof before merge, or accept the isolated test-store proof under the repo's no-live-Keychain-prompt policy?
• Rationale: The patch changes OAuth credential cache behavior and repository policy discourages live prompt-producing validation, so the remaining blocker is proof sufficiency rather than a line-level code defect.
• Likely owner: steipete — steipete is the strongest current release and Keychain prompt policy owner in the relevant history.
• Options:
  • Require affected-app proof (recommended): Ask for redacted after-fix runtime evidence, such as unified logs across a refresh window showing no CodexBar Keychain activity for oauth.claude while .never is set.
  • Accept isolated proof: A maintainer can explicitly override the proof gate if isolated test-keychain coverage is considered sufficient for this prompt path.
  • Pause for maintainer machine validation: Keep the PR open until someone with a safe affected setup can validate the no-prompt behavior without exposing private credentials.

Security
Cleared: No concrete security or supply-chain regression was found; the remaining auth-sensitive concern is proof and upgrade behavior rather than a discrete security defect.

Review details

Best possible solution:

Land the focused cache-bypass fix after redacted affected-app runtime proof or an explicit maintainer proof override confirms that .never avoids oauth.claude Keychain touches without regressing Claude auth recovery.

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

Yes, source-backed: the linked issue includes affected-machine diagnostics, and current main reads the CodexBar-owned Claude OAuth cache before file fallback even when prompt mode is .never. I did not run live Keychain validation because AGENTS.md forbids unrequested prompt-producing checks.

Is this the best way to solve the issue?

Yes for the code shape: gating the Claude-specific CodexBar cache above KeychainCacheStore is narrower than changing the global cache layer, and the PR covers stale-cache tombstone paths. The merge still needs real behavior proof or a maintainer proof override for the auth-sensitive runtime path.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 3b039d15e125.

Label changes

Label justifications:

• P2: This is a normal-priority Claude auth-provider prompt-loop fix with limited blast radius and no confirmed data loss or core runtime outage.
• merge-risk: 🚨 auth-provider: The PR changes Claude OAuth credential cache read/write/clear behavior and could affect token loading or recovery for existing .never prompt-mode users.
• rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🦞 diamond lobster.
• 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 includes copied focused Swift test output and simulated logs from isolated test-store/test-keychain runs, but no after-fix runtime proof from an affected CodexBar app setup. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence reviewed

What I checked:

• Repository policy applied: AGENTS.md was read fully; its guidance to avoid unrequested macOS Keychain-prompt validation is directly relevant to judging proof and validation for this PR. (AGENTS.md:25, 3b039d15e125)
• Current main cache path: Current main reads the CodexBar-owned Claude OAuth Keychain cache before falling back to the credentials file, matching the linked issue's diagnosed prompt path. (Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials.swift:228, 3b039d15e125)
• PR head cache bypass: At PR head, loadRecord, prompt recovery, and cached-credential checks route through shouldUseCodexBarOAuthKeychainCache, which is false when stored prompt mode is .never. (Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials.swift:205, ab8301504444)
• Pending-clear behavior: The PR adds a pending-clear tombstone, retries Keychain cache deletion, treats an unflushed tombstone as temporarily unavailable, and keeps the marker when replacement storage fails. (Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials.swift:2137, ab8301504444)
• Regression coverage: The added suite covers never-mode file preference, skipped cache writes, idempotent invalidation, memory-cache bypass, pending-clear retry, and failed replacement-store handling using isolated test stores. (Tests/CodexBarTests/ClaudeOAuthCredentialsStoreNeverPromptCacheTests.swift:27, ab8301504444)
• Linked issue diagnosis: The canonical linked issue includes affected-machine diagnostics tying the v0.41.0 periodic prompt loop to .never prompt mode, five-minute refresh cadence, and oauth.claude cache access before file fallback.

Likely related people:

• steipete: Current blame on the shipped Claude OAuth cache path collapses to the v0.41.0 release commit, and recent history shows Keychain prompt trust and cache-store work in this area. (role: recent area contributor and release owner; confidence: high; commits: a83a83fa4131, 8b417a066cb1, 24fe798f3769; files: Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials.swift, Sources/CodexBarCore/KeychainCacheStore.swift, docs/keychain-prompts.md)
• ratulsarna: GitHub path history shows the densest earlier work on Claude OAuth prompt policy, delegated recovery, security-CLI reading, and keychain-cache lock handling. (role: feature-history owner; confidence: high; commits: 0e7da356a37f, 5b52899ea62a, 8736b0a4460c; files: Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials.swift, Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials+SecurityCLIReader.swift, Sources/CodexBarCore/KeychainCacheStore.swift)
• ss251: Recent main history ties ss251 to Claude OAuth cache/history correctness around account switching and stale credential ownership. (role: recent adjacent contributor; confidence: medium; commits: 9a0028442e35; files: Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials.swift)
• Yuxin-Qiao: Although also the PR author, prior merged history ties Yuxin-Qiao to background Claude delegated-refresh and MCP-only keychain-touch suppression adjacent to this prompt loop. (role: recent adjacent contributor; confidence: medium; commits: a27c9b094e61; files: Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials.swift, Sources/CodexBarCore/Providers/Claude/ClaudeOAuth/ClaudeOAuthCredentials+SecurityCLIReader.swift, Sources/CodexBarCore/Providers/Claude/ClaudeUsageFetcher.swift)

What the crustacean ranks mean

• 🦀 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.

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.

Review history (4 earlier review cycles)

• reviewed 2026-07-09T16:42:57.091Z sha 8a42ab7 :: needs real behavior proof before merge. :: [P1] Keep skipped cache invalidation pending
• reviewed 2026-07-09T16:56:43.438Z sha f37327e :: needs real behavior proof before merge. :: none
• reviewed 2026-07-09T17:47:21.358Z sha 28f7c02 :: needs real behavior proof before merge. :: none
• reviewed 2026-07-09T18:05:23.456Z sha ab83015 :: needs real behavior proof before merge. :: none

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8a42ab76ef

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f37327eabd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 38a399deac

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3c340b40d3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

[Yuxin-Qiao]

Addressed all seven @chatgpt-codex-connector P2 inline comments in commits 3c340b40–28f7c02b (pending-clear tombstone, memory-cache bypass, fingerprint idempotency, MCP-only effectiveMode, ownership-test update, locked-keychain retry, store-failure handling).

Updated the PR body with redacted macOS behavior proof from isolated test-keychain runs (no live Keychain UI per AGENTS.md).

@clawsweeper re-review

[clawsweeper]

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

[Yuxin-Qiao]

Live macOS behavior proof (#1991 / #2025)

Reproduced on a local machine against the installed Homebrew/Sparkle app 0.41.0 (100), not just isolated Swift tests.

Setup

• defaults write com.steipete.codexbar claudeOAuthKeychainPromptMode never
• Seeded CodexBar-owned Keychain cache item service=com.steipete.codexbar.cache, account=oauth.claude
• Ensured ~/.claude/.credentials.json fingerprint was already established (so refresh does not immediately clear the cache on first invalidation)
• Stopped the menu-bar app to avoid concurrent Keychain access, then invoked:

/Applications/CodexBar.app/Contents/Helpers/CodexBarCLI usage --provider claude --source oauth

Result

• CodexBarCLI hung for >=12s until killed, consistent with a visible macOS Keychain password prompt while never mode is active.
• This matches the reported periodic refresh loop: background OAuth refresh still touches the CodexBar-owned oauth.claude cache even when direct Claude Code Keychain reads are blocked.

Repro helper

Added a small helper script on my branch:

./Scripts/repro_keychain_issues.sh 2025

It seeds the cache, switches to never, and watches for the expected hang.

Notes

• I intentionally did not rely on live Keychain UI during CI/tests; this comment is the maintainer-requested real-setup proof.
• Related fresh report: Always asking for Mac Authentication #2025 (same symptom, likely duplicate of v0.41.0: macOS Keychain password prompt repeats every few minutes #1991).

[Yuxin-Qiao]

@clawsweeper re-review — added live macOS repro proof in the previous comment and pushed Scripts/repro_keychain_issues.sh in 4a41e582.

[clawsweeper]

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4a41e5823d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3cb19cc030

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[clawsweeper]

ClawSweeper status: review started.

I am starting a fresh review of this pull request: Fix Claude never-prompt Keychain cache bypass (#1991) This is item 1/1 in the current shard. Shard 0/1.

This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking.

Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted.

[clawsweeper]

ClawSweeper status: review started.

I am starting a fresh review of this pull request: Fix Claude never-prompt Keychain cache bypass (#1991) This is item 1/1 in the current shard. Shard 0/1.

This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking.

Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted.