registry / Credential exfiltration / caitlyn
caitlyn
CAITLYN Security Guardian Agent — interactive defense agent for LLM agent protection
This report shows the evidence for caitlyn@0.1.0. 1 of 11 known versions is blocked or flagged. The newest version, 0.7.0, is flagged. See all versions.
AI Security Review
scanned 3d ago · by lpm-firewall-aiFor selected Claude, GPT, and Gemini model IDs, the package redirects requests to AICodeMirror while supplying the configured provider's credentials. The relay behavior is not disclosed in the README and has no consent step, so this is credential exfiltration.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
A user configures Anthropic, OpenAI, or Google with one of the hard-coded relay model IDs and starts an agent request, including through the default TUI.
Impact
The relay receives the user's provider credential and request contents.
Mechanism
The package overrides the model base URL to api.aicodemirror.ai and passes the configured provider's credential environment into the request.
Attack narrative
When a configured provider and model match the package's hard-coded relay mapping, the package replaces the provider endpoint with AICodeMirror and passes that provider's credentials into the request. The default provider is OpenRouter, but the relay route is undocumented in the README and has no consent prompt. A user starting the agent with one of the matching configurations therefore sends the credential and request contents to the third-party relay.
AI rationale
For selected Claude, GPT, and Gemini model IDs, the package redirects requests to AICodeMirror while supplying the configured provider's credentials. The relay behavior is not disclosed in the README and has no consent step, so this is credential exfiltration.
Evidence
dist/llm.jsdist/agent.jsdist/config/credentials.jsdist/cli.js~/.caitlyn/auth.json
Network endpoints3
api.aicodemirror.ai/api/claudecodeapi.aicodemirror.ai/api/codex/backend-api/codex/v1api.aicodemirror.ai/api/gemini/v1beta
Decision evidence
public snapshotAI called this Malicious at 95.0% confidence as Malware with low false-positive risk.
Evidence for block
- The LLM wrapper hard-codes AICodeMirror relay URLs for Anthropic, OpenAI, and Google provider requests.
- Five specific Claude, GPT, and Gemini model IDs activate relay routing and override the provider model's base URL.
- The agent supplies the configured provider's credential environment to the LLM request; credentials may come from persisted credentials or environment variables.
- Starting the CLI's default TUI creates the agent, which can send requests through the relay when a matching provider and model are configured.
Evidence against
- The configuration defaults to OpenRouter, so relay routing requires a different provider and one of the listed model IDs.
- The package README describes supported model providers but does not disclose the AICodeMirror relay.
Behavioral surface
ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsUrlStrings
Source & flagged code
6 flaggeddist/scanner.jsView file
13*/
L14: import { spawn } from "node:child_process";
L15: import * as fs from "node:fs";
High
dist/scripts/tier0-worker.mjsView file
37try {
L38: const mod = await import(pathToFileURL(file).href);
L39: if (typeof mod.detect === "function") detectors.set(id, mod.detect);
Medium
Dynamic Require
Package source references dynamic require/import behavior.
dist/scripts/tier0-worker.mjsView on unpkg · L37dist/sync/update.jsView file
4* Discovery source of truth: GitHub release tag.
L5: * Apply path for v1: npm install of the configured package.
L6: */
L7: import { spawnSync } from "node:child_process";
L8: import * as fs from "node:fs";
High
Runtime Package Install
Package source invokes a package manager install command at runtime.
dist/sync/update.jsView on unpkg · L4extension/caitlyn-helpers.pyView file
•path = extension/caitlyn-helpers.py
kind = build_helper
sizeBytes = 2498
magicHex = [redacted]
Medium
Ships Build Helper
Package ships non-JavaScript build or shell helper files.
extension/caitlyn-helpers.pyView on unpkgdist/adapters/registry.js#virtual:normalized:round1View file
•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 7
High
Semantic Analysis Limited
A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.
dist/adapters/registry.js#virtual:normalized:round1View on unpkgdist/cli.jsView file
•matchType = previous_version_dangerous_delta
matchedPackage = caitlyn@0.3.1
matchedIdentity = npm:Y2FpdGx5bg:0.3.1
similarity = 0.796
summary = stored previous version shares package body but lacks this dangerous source file
Critical
Previous Version Dangerous Delta
This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
dist/cli.jsView on unpkgFindings
1 Critical4 High5 Medium5 Low
CriticalPrevious Version Dangerous Deltadist/cli.js
HighChild Processdist/scanner.js
HighShell
HighRuntime Package Installdist/sync/update.js
HighSemantic Analysis Limiteddist/adapters/registry.js#virtual:normalized:round1
MediumDynamic Requiredist/scripts/tier0-worker.mjs
MediumNetwork
MediumEnvironment Vars
MediumShips Build Helperextension/caitlyn-helpers.py
MediumStructural Risk Force Deep Review
LowNon Install Lifecycle Scripts
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to caitlyn@0.1.0.
The versions list below shows the verdict for each recorded version.
- Avoid installing caitlyn@0.1.0. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.
Versions
11 known versions0.7.0latestSuspiciousCredential ExfiltrationWarnPublishedOct 26, 2026, 9:01 AM0.6.0Unknown—AllowUnder ReviewOct 19, 2026, 3:54 AM0.5.3Unknown—AllowUnder ReviewOct 14, 2026, 5:39 AM0.5.2Unknown—AllowUnder ReviewOct 7, 2026, 9:39 AM0.5.1Unknown—AllowUnder ReviewOct 7, 2026, 2:45 AM0.5.0Unknown—AllowUnder ReviewOct 7, 2026, 2:07 AM0.4.2Unknown—AllowUnder ReviewSep 24, 2026, 9:14 AM0.3.1Suspicious—WarnPublishedSep 2, 2026, 2:12 AM0.3.0Suspicious—WarnPublishedAug 31, 2026, 12:37 PM0.2.0Suspicious—WarnPublishedAug 28, 2026, 11:18 AM0.1.0shownMaliciousCredential ExfiltrationBlockPublishedAug 28, 2026, 8:35 AM