registry / @rallycry/conveyor-agent / 11.0.2
@rallycry/conveyor-agent@11.0.2
Conveyor Agent Runner v10 - PTY harness for the task chat (SDK harness for audit/project-chat). Agent-as-User architecture with BaseService patterns. Works locally too.
AI Security Review
scanned 12d ago · by lpm-firewall-aiThe manifest has no npm install lifecycle hook, and the inspected credential changes support the package’s agent runtime. The scanner’s credential and persistence hints do not establish an unrelated credential recipient or a concrete attack.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review; previous stored version diff introduced dangerous source
Trigger
The runtime boot command configures the agent environment; upload behavior occurs when the workbench requests a file upload.
Impact
The inspected behavior supports agent operation and user-requested file sharing; it does not establish credential theft or another attack.
Mechanism
Boot code copies credentials to a pod-local file and links the Claude config path to it. Upload code reads a selected workspace file and PUTs it to the upload URL returned by the workbench connection.
AI rationale
The manifest has no npm install lifecycle hook, and the inspected credential changes support the package’s agent runtime. The scanner’s credential and persistence hints do not establish an unrelated credential recipient or a concrete attack.
Decision evidence
public snapshotAI called this Clean at 89.0% confidence as Benign with low false-positive risk.
Evidence for AI clean decision
- package.json defines prepublishOnly as tsup and has no preinstall, install, or postinstall hook.
- The boot code copies Claude credentials into a pod-local file and repoints the package’s shared Claude credential path to it; this is local agent setup.
- Claude credentials are written from environment-provided OAuth material to the Claude config directory; the inspected code does not show those credentials being sent to an unrelated endpoint.
- The flagged chunks implement package functions for git credential setup, workbench framing, and an environment-configured OOM watchdog.
- The inspected upload code sends selected workspace file bytes to a URL returned by the workbench upload request.
Evidence against
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShellWebSocket
HighEntropyStringsPossibleObfuscation
Source & flagged code
12 flagged · loading sourcedist/chunk-QETNCXIR.jsView file
129try {
L130: parsed = JSON.parse(payloadJson);
L131: } catch {
...
L544: // ── Environment setup / start command ─────────────────────────────────
L545: z3.object({ type: z3.literal("setup_output"), stream: z3.string(), data: z3.string() }).passthrough(),
L546: z3.object({
...
L2094: "xpack.profiling.enabled": "false",
L2095: "ingest.geoip.downloader.enabled": "false",
L2096: ES_JAVA_OPTS: "-Xms512m -Xmx512m"
...
L2109: connectionEnv: {
L2110: ELASTICSEARCH_URL: "http://elasticsearch:9200"
L2111: },
Critical
Credential Exfiltration
Source appears to send environment or credential material to an external endpoint.
dist/chunk-QETNCXIR.jsView on unpkg · L129129Trigger-reachable chain: manifest.main -> dist/index.js -> dist/chunk-QETNCXIR.js
L129: try {
L130: parsed = JSON.parse(payloadJson);
L131: } catch {
...
L544: // ── Environment setup / start command ─────────────────────────────────
L545: z3.object({ type: z3.literal("setup_output"), stream: z3.string(), data: z3.string() }).passthrough(),
L546: z3.object({
...
L2094: "xpack.profiling.enabled": "false",
L2095: "ingest.geoip.downloader.enabled": "false",
L2096: ES_JAVA_OPTS: "-Xms512m -Xmx512m"
...
L2109: connectionEnv: {
L2110: ELASTICSEARCH_URL: "http://elasticsearch:9200"
L2111: },
Critical
Trigger Reachable Dangerous Capability
A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.
dist/chunk-QETNCXIR.jsView on unpkg · L129Findings
4 Critical7 High4 Medium5 Low
CriticalCredential Exfiltrationdist/chunk-QETNCXIR.js
CriticalPersistence Backdoordist/boot-PKQ2I66D.js
CriticalTrigger Reachable Dangerous Capabilitydist/chunk-QETNCXIR.js
CriticalPrevious Version Dangerous Deltadist/boot-PKQ2I66D.js
HighTrigger Reachable Credential Exfiltrationdist/chunk-QETNCXIR.js
HighTrigger Reachable Persistencedist/boot-PKQ2I66D.js
HighSemantic Analysis Limiteddist/chunk-QETNCXIR.js
HighKnown Malware Source Similaritydist/chunk-GL2DIQEQ.js
HighKnown Malware Source Similaritydist/chunk-EXQ6AHOY.js
HighKnown Malware Source Similaritydist/chunk-6W6UZ4SJ.js
HighKnown Malware Source Similarity
Affected versions and remediation
This report applies to @rallycry/conveyor-agent@11.0.2.
See all versions of @rallycry/conveyor-agent for other recorded verdicts.
- Review the evidence and your use of @rallycry/conveyor-agent@11.0.2 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.