Keep Claude CLI sessions warm by obviyus · Pull Request #69679 · openclaw/openclaw

obviyus · 2026-04-21T10:34:50Z

Summary

Keep one Claude stdio process warm per OpenClaw session.
Close idle Claude processes after 10 minutes, then resume with the stored Claude session id.
Add config/schema/docs plus regression coverage for reuse and idle cleanup.

Tests

pnpm test src/agents/cli-runner.spawn.test.ts src/agents/cli-output.test.ts
pnpm check:changed
Live smoke: prepared OpenClaw Claude stdio path reused one Claude session id across two turns.

aisle-research-bot · 2026-04-21T10:34:58Z

🔒 Aisle Security Analysis

We found 3 potential security issue(s) in this PR:

#	Severity	Title
1	🟡 Medium	CLI session reuse ignores OpenClaw MCP loopback port, enabling session binding confusion
2	🟡 Medium	Unbounded stdout/stderr buffering after Claude live session is marked closing (memory DoS)
3	🟡 Medium	Global Claude live-session cap allows cross-tenant denial of service

1. 🟡 CLI session reuse ignores OpenClaw MCP loopback port, enabling session binding confusion

Property	Value
Severity	Medium
CWE	CWE-346
Location	`src/agents/cli-runner/bundle-mcp.ts:273-280`

Description

mcpResumeHash intentionally canonicalizes the OpenClaw MCP URL by replacing the loopback port with a constant placeholder. resolveCliSessionReuse() then prefers this resume hash over the full mcpConfigHash.

This is a session-binding weakening/regression:

Input: MCP endpoint URL is part of the MCP config (mcpServers.openclaw.url).
Canonicalization: normalizeOpenClawLoopbackUrl() strips the port for http://127.0.0.1|localhost|[::1]:<port>/mcp.
Decision: resolveCliSessionReuse() will reuse a stored Claude session ID as long as mcpResumeHash matches, even if the actual MCP endpoint port changed.

Impact:

If the OpenClaw MCP server restarts on a different port and that port is (accidentally or maliciously) bound by a different local process speaking MCP, the CLI may resume a prior Claude session (with prior conversation/tool expectations) against a meaningfully different tool backend.
Previously, the raw config hash would invalidate the session on any URL change (including port), reducing the chance of resuming across endpoint changes.

Vulnerable code:

/^(http:\/\/(?:127\.0\.0\.1|localhost|\[::1\])):\d+(\/mcp)$/.exec(value.trim())
...
return `${match[1]}:<openclaw-loopback>${match[2]}`;

Recommendation

Do not treat the MCP port as non-semantic for session reuse unless you can cryptographically bind the resumed Claude session to the same local MCP server instance.

Safer options:

Keep the port in the resume hash (revert this normalization) and instead persist the originally chosen port for OpenClaw so subsequent runs reuse the same port when possible.
Bind to server identity: include a stable server identifier in the resume hash (e.g., a generated instance ID or public key fingerprint) and require the MCP server to prove possession during handshake.

Example (simple mitigation: include port):

function canonicalizeBundleMcpConfigForResume(config: BundleMcpConfig): BundleMcpConfig {
  return { mcpServers: sortJsonValue(config.mcpServers) as BundleMcpConfig["mcpServers"] };
}

If you must ignore ephemeral ports, add an explicit allowlist flag (e.g., allowSessionReuseAcrossPortChanges: true) and default it to false so users must opt in.

2. 🟡 Unbounded stdout/stderr buffering after Claude live session is marked closing (memory DoS)

Property	Value
Severity	Medium
CWE	CWE-400
Location	`src/agents/cli-runner/claude-live-session.ts:324-349`

Description

The Claude CLI live session continues to append to stdoutBuffer/stderr even after the session has been closed (session.closing=true).

closeLiveSession() returns immediately when session.closing is already true.
handleClaudeStdout()/onStderr keep appending data and rely on closeLiveSession() to abort when limits are exceeded.
Once session.closing is true, further limit checks still call closeLiveSession(), but it becomes a no-op, so buffers can grow without bound until the child process actually exits.

This allows a misbehaving/compromised claude CLI process (or a local wrapper script invoked as the CLI) to stream unlimited output after cancellation/abort and cause memory exhaustion (denial of service).

Recommendation

Stop processing output once a session is closing, and/or hard-cap buffers even in the closing state.

Suggested fix (minimal):

function handleClaudeStdout(session: ClaudeLiveSession, chunk: string) {
  if (session.closing) return;
  ...
}

// in onStderr
if (session.closing) return;

Additionally, consider:

Clearing buffers on close to release memory: session.stdoutBuffer = ""; session.stderr = "";
Detaching stdout/stderr handlers (if supported by the supervisor) or ensuring managedRun.cancel() promptly destroys the streams.
Keeping the limit enforcement independent of closeLiveSession() (e.g., truncate buffers) so it still works when already closing.

3. 🟡 Global Claude live-session cap allows cross-tenant denial of service

Property	Value
Severity	Medium
CWE	CWE-400
Location	`src/agents/cli-runner/claude-live-session.ts:54-756`

Description

The Claude CLI live-session implementation enforces a single, process-wide cap (CLAUDE_LIVE_MAX_SESSIONS = 16) across all live sessions and pending session creations.

Session entries are stored in global in-memory maps (liveSessions, liveSessionCreates).
ensureLiveSessionCapacity() rejects new sessions once liveSessions.size + liveSessionCreates.size reaches the global cap.
The session key is per agent/account/session (buildClaudeLiveKey() includes agentAccountId, agentId, authProfileId, sessionId, sessionKey), but the capacity control is not scoped by tenant/account.

In a multi-tenant deployment (e.g., a shared gateway/service process handling multiple users/accounts), a single tenant can create/hold 16 sessions (or keep 16 creates pending) and cause other tenants to receive a FailoverError with reason rate_limit (“Too many Claude CLI live sessions are active.”), effectively producing a cross-user denial-of-service.

Vulnerable code:

const CLAUDE_LIVE_MAX_SESSIONS = 16;
...
if (
  liveSessions.has(key) ||
  liveSessionCreates.has(key) ||
  liveSessions.size + liveSessionCreates.size < CLAUDE_LIVE_MAX_SESSIONS
) {
  return;
}
...
throw new FailoverError("Too many Claude CLI live sessions are active.", { reason: "rate_limit", ... });

Recommendation

Scope capacity controls to the resource owner and add fair eviction:

Enforce caps per tenant/account (e.g., per agentAccountId and/or authProfileId) rather than globally.
Optionally keep a small global cap as a final safety limit, but ensure per-tenant fairness.
Implement true LRU/idle eviction (track lastUsedAtMs and evict the least-recently-used idle session), and consider closing sessions that exceed a maximum lifetime.
Consider admission control: refuse/queue additional sessions per tenant instead of failing other tenants.

Example approach (sketch):

type TenantKey = string; // e.g. sha256(agentAccountId + authProfileId)
const tenantSessions = new Map<TenantKey, Set<string>>();
const MAX_SESSIONS_PER_TENANT = 4;

function ensureTenantCapacity(tenant: TenantKey) {
  const set = tenantSessions.get(tenant) ?? new Set();
  if (set.size >= MAX_SESSIONS_PER_TENANT) {
    throw new FailoverError("Too many live sessions for this account.", { reason: "rate_limit", ... });
  }
}

This prevents one tenant from exhausting shared process resources and causing cross-tenant service degradation.

Analyzed PR: #69679 at commit e23d9d2

_{Last updated on: 2026-04-22T08:13:12Z}

greptile-apps · 2026-04-21T10:38:51Z

Greptile Summary

This PR introduces a warm-process optimization for the claude-cli backend: one Claude stdio process is kept alive per OpenClaw session, follow-up turns are sent as stream-json stdin messages, and the process is cancelled after 10 minutes of idle (with the stored Claude session id used to reconnect on next use). The implementation adds claude-live-session.ts for process lifecycle management, updates execute.ts to dispatch into the new path, extends config schema/types, and ships two regression tests verifying session reuse and idle cleanup.

Confidence Score: 5/5

Safe to merge; all remaining findings are P2 style/minor correctness notes that don't block the primary user path.

No P0 or P1 issues. The two logic comments (stale noOutputTimeoutMs and accumulated stderr) are P2 because noOutputTimeoutMs stays constant within a session in practice (same backend fingerprint), and the stderr mixing only matters on unexpected process exit after multiple turns. The unused streamingParser allocation is a style issue only.

src/agents/cli-runner/claude-live-session.ts — see noOutputTimeoutMs and stderr comments.

Comments Outside Diff (1)

src/agents/cli-runner/execute.ts, line 334-360 (link)

streamingParser allocated but never used on the live-session path

streamingParser is constructed unconditionally when backend.output === "jsonl", then its truthiness gates entry into the live-session branch, but the object itself is never pushed to or finished — runClaudeLiveSessionTurn creates its own internal parser. The null-check pattern is clear as a guard, but consider gating the allocation (const hasJsonlOutput = backend.output === "jsonl") to avoid the dead object, or at least a short comment noting the intentional discard.

Prompt To Fix With AI

This is a comment left during a code review.
Path: src/agents/cli-runner/execute.ts
Line: 334-360

Comment:
**`streamingParser` allocated but never used on the live-session path**

`streamingParser` is constructed unconditionally when `backend.output === "jsonl"`, then its truthiness gates entry into the live-session branch, but the object itself is never pushed to or finished — `runClaudeLiveSessionTurn` creates its own internal parser. The null-check pattern is clear as a guard, but consider gating the allocation (`const hasJsonlOutput = backend.output === "jsonl"`) to avoid the dead object, or at least a short comment noting the intentional discard.

How can I resolve this? If you propose a fix, please make it concise.

Prompt To Fix All With AI

This is a comment left during a code review.
Path: src/agents/cli-runner/claude-live-session.ts
Line: 411-433

Comment:
**Stale `noOutputTimeoutMs` in reused sessions**

The `onStdout` and `onStderr` handlers closed over `params.noOutputTimeoutMs` at session-creation time. When a second turn arrives with a different `timeoutMs` (and thus a different computed `noOutputTimeoutMs`), the per-turn timer created in `createTurn` uses the correct value — but every subsequent stdout/stderr chunk resets that timer back to the first turn's timeout via `resetNoOutputTimer(session, params.noOutputTimeoutMs)`.

In practice, the fingerprint check guarantees the backend config is identical across turns in the same session, so the watchdog config rarely differs. However, `params.timeoutMs` (from `context.params.timeoutMs`) is per-request and could legitimately vary, causing the watchdog to fire too early or too late for later turns.

A clean fix is to store `noOutputTimeoutMs` on `ClaudeLiveSession` and update it at the start of each turn in `runClaudeLiveSessionTurn`.

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: src/agents/cli-runner/claude-live-session.ts
Line: 333-370

Comment:
**`session.stderr` accumulates across turns**

`session.stderr` is never reset between turns — stderr from turn 1 is still present when turn 2 fails. If the process exits unexpectedly mid-turn-2, `handleClaudeExit` will call `extractCliErrorMessage(session.stderr)` on the combined multi-turn stderr, which can surface stale or misleading error context. Consider clearing `session.stderr` at the beginning of each new turn (e.g., inside `runClaudeLiveSessionTurn` before `writeTurnInput`).

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: src/agents/cli-runner/execute.ts
Line: 334-360

Comment:
**`streamingParser` allocated but never used on the live-session path**

`streamingParser` is constructed unconditionally when `backend.output === "jsonl"`, then its truthiness gates entry into the live-session branch, but the object itself is never pushed to or finished — `runClaudeLiveSessionTurn` creates its own internal parser. The null-check pattern is clear as a guard, but consider gating the allocation (`const hasJsonlOutput = backend.output === "jsonl"`) to avoid the dead object, or at least a short comment noting the intentional discard.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "test(cli): cover claude live session reu..." | Re-trigger Greptile}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ff45a2a337

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: ea62254fb4

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: 4b0546dbab

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: 6dbc6bf198

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: 808675ec1a

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: 012f390901

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: f5b1959511

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: 40f769dfeb

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: e6b40b2a7f

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: 704828f469

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 · 2026-04-21T14:33:47Z

+  cleanup: () => Promise<void>;
+}): Promise<ClaudeLiveRunResult> {
+  const key = buildClaudeLiveKey(params.context);
+  const useResume = params.args.includes("--resume") || params.args.includes("-r");


Detect inline --resume values before deciding live reuse

runClaudeLiveSessionTurn only treats resume mode as active when args contains a standalone --resume/-r token. If a backend uses resumeArgs like "--resume={sessionId}" (single argv item, which is allowed by the {sessionId} placeholder contract), useResume is false here, so the turn is treated as a fresh run: buildClaudeLiveArgs re-adds first-turn system prompt behavior and the resumeCapable && !useResume restart path tears down warm sessions unnecessarily. That breaks resumed-session continuity for valid inline resume flag configurations.

Useful? React with 👍 / 👎.

obviyus · 2026-04-22T08:14:27Z

Landed on main.

Landed source commit: e23d9d2
Merge commit: 81ca7bc

* feat(cli): keep claude cli sessions warm * test(cli): cover claude live session reuse * fix(cli): harden claude live session reuse * fix(cli): redact mcp session key logs * fix(cli): bound claude live session turns * fix(cli): reuse claude live sessions on resume * refactor(cli): canonicalize claude live argv * fix(cli): preserve claude live resume state * fix(cli): close dead claude live sessions * fix(cli): serialize claude live session creates * fix(cli): count pending claude live sessions * fix(cli): tighten claude live resume abort * fix(cli): reject closed claude live sessions * fix(cli): refresh claude live fingerprints * fix(cli): stabilize MCP resume hash * fix: preserve claude live inline resume (openclaw#69679) --------- Co-authored-by: Frank Yang <[email protected]>

…resume OpenClaw v2026.4.22 (released 2026-04-23) introduced native warm claude-cli stdio sessions and automatic resume from the stored Claude session id after Gateway restarts or idle exits, via openclaw/openclaw#69679 "Keep Claude CLI sessions warm". This covers the primary use case of boot-resume, so the standalone skill is no longer needed. This commit adds prominent deprecation banners to README.md (English and Chinese sections) and SKILL.md, pointing readers to the upstream PR. The repository will be archived after this commit. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

* feat(cli): keep claude cli sessions warm * test(cli): cover claude live session reuse * fix(cli): harden claude live session reuse * fix(cli): redact mcp session key logs * fix(cli): bound claude live session turns * fix(cli): reuse claude live sessions on resume * refactor(cli): canonicalize claude live argv * fix(cli): preserve claude live resume state * fix(cli): close dead claude live sessions * fix(cli): serialize claude live session creates * fix(cli): count pending claude live sessions * fix(cli): tighten claude live resume abort * fix(cli): reject closed claude live sessions * fix(cli): refresh claude live fingerprints * fix(cli): stabilize MCP resume hash * fix: preserve claude live inline resume (openclaw#69679) --------- Co-authored-by: Frank Yang <[email protected]>

openclaw-barnacle Bot added docs Improvements or additions to documentation gateway Gateway runtime agents Agent runtime and tooling extensions: anthropic size: L maintainer Maintainer-authored PR labels Apr 21, 2026

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts Outdated

Comment thread src/agents/cli-runner/claude-live-session.ts Outdated

greptile-apps Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts

Comment thread src/agents/cli-runner/claude-live-session.ts

openclaw-barnacle Bot added size: XL and removed size: L labels Apr 21, 2026

obviyus self-assigned this Apr 21, 2026

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts

Comment thread src/agents/cli-runner/claude-live-session.ts Outdated

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts Outdated

Comment thread src/agents/cli-runner/execute.ts

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts

Comment thread src/agents/cli-runner/claude-live-session.ts

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts Outdated

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts Outdated

Comment thread src/agents/cli-runner/claude-live-session.ts

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

Comment thread src/agents/cli-runner/claude-live-session.ts

Comment thread src/agents/cli-runner/claude-live-session.ts

chatgpt-codex-connector Bot reviewed Apr 21, 2026

View reviewed changes

obviyus added 8 commits April 22, 2026 13:31

feat(cli): keep claude cli sessions warm

faf58b9

test(cli): cover claude live session reuse

028b755

fix(cli): harden claude live session reuse

6ca280a

fix(cli): redact mcp session key logs

dd281d5

fix(cli): bound claude live session turns

23298cb

fix(cli): reuse claude live sessions on resume

d7e8b0d

refactor(cli): canonicalize claude live argv

cecb5ed

fix(cli): preserve claude live resume state

f273bc4

obviyus force-pushed the feat/claude-live-session branch from 490041f to e23d9d2 Compare April 22, 2026 08:09

obviyus merged commit 81ca7bc into openclaw:main Apr 22, 2026
93 checks passed

github-actions Bot mentioned this pull request Apr 22, 2026

📡 Upstream Digest — 2026-04-22 09:12 UTC curtismercier/openclaw-mods#643

Open

perlowja mentioned this pull request Apr 23, 2026

[Bug] Agent tool-use regressed for all providers between v2026.4.20 and main — 0/6 models invoke tools on a cold-start prompt; multiple model families converge on fabricated "Exec is blocked on this session" phrase #70456

Closed

poetmiku mentioned this pull request Apr 24, 2026

[Bug]: CLI sessions: Gemini CLI not covered by Claude CLI session persistence fixes (#69679 / #70106 / #70132) — gateway restart still mints a fresh conversation #70973

Closed

chacher86 mentioned this pull request Apr 25, 2026

claude-cli live session: hardcoded 256KB stdout buffer causes spurious "Claude CLI stdout buffer exceeded limit." failures #71793

Closed

clawsweeper Bot mentioned this pull request May 12, 2026

feat: add agents.defaults.systemPromptFile config option #41807

Closed

18 tasks

This was referenced May 21, 2026

feat: raise CLI watchdog no-output timeout caps (3 min → 30 min) #41022

Closed

feat: raise (or remove) 3-minute no-output watchdog cap on CLI requests #40982

Open

clawsweeper Bot mentioned this pull request Jun 14, 2026

fix(claude-cli): fork session on resume to avoid concurrent collisions #77148

Closed

25 tasks

clawsweeper Bot mentioned this pull request Jun 26, 2026

[Bug]: EPIPE on claude-cli live-session spawn (Docker, Unraid, OAuth) #77178

Closed

This was referenced Jul 6, 2026

fix(agents): guard CLI MCP resume reuse #90979

Open

fix(cli-runner): sweep orphaned bundle MCP temp dirs at gateway startup #105643

Open

clawsweeper Bot mentioned this pull request Jul 18, 2026

CLI timeouts can replay active work and give misleading guidance #110340

Closed

Uh oh!

Uh oh!

Conversation

obviyus commented Apr 21, 2026

Uh oh!

aisle-research-bot Bot commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔒 Aisle Security Analysis

Description

Recommendation

Description

Recommendation

Description

Recommendation

Uh oh!

greptile-apps Bot commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Comments Outside Diff (1)

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Apr 21, 2026

Choose a reason for hiding this comment

aisle-research-bot Bot commented Apr 21, 2026 •

edited

Loading

greptile-apps Bot commented Apr 21, 2026 •

edited

Loading