wildcrane-runner@0.4.0
WildCrane's Local Runner — connects this computer to WildCrane and runs its jobs on your local Claude Code, Codex, OpenCode or Cursor.
AI Security Review
scanned 5d ago · by lpm-firewall-aiThe package’s network and process activity matches its disclosed purpose: connecting a user’s local agent tools to WildCrane tasks. The flagged credential and shell references did not reveal automatic exfiltration or unrelated execution.
Decision evidence
public snapshot- The package describes itself as connecting to WildCrane and running its jobs through local Claude Code, Codex, OpenCode, or Cursor.
- The README says the runner signs in through a browser on first start and updates to the version WildCrane names.
- The launcher verifies downloaded runner tarballs against the npm registry’s SHA-512 integrity value before installing them.
- The runner passes the environment to the selected local agent harness to carry out the requested task.
- The credential-related references are provider settings and environment passed to local agent tools; I found no flow sending discovered credentials to an unrelated host.
- The curl-to-shell strings are installer commands for supported agent tools, not automatic startup or install actions.
- The prepare script invokes a build command; no install-time data collection or unrelated network contact was identified.
Source & flagged code
9 flaggedSource appears to send environment or credential material to an external endpoint.
dist/akb/dist/kanban.mjsView on unpkg · L58Package source references child process execution.
dist/akb/dist/kanban.mjsView on unpkg · L1168Findings
Affected versions and remediation
This report applies to wildcrane-runner@0.4.0.
See all versions of wildcrane-runner for other recorded verdicts.
- Review the evidence and your use of wildcrane-runner@0.4.0 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.