@cursor/sdk@1.0.22
TypeScript SDK for Cursor agents.
AI Security Review
scanned 4mo ago · by lpm-firewall-aiStatic inspection shows a legitimate Cursor agent SDK: network, env, filesystem, and agent tool primitives are tied to explicit SDK calls and local/cloud agent functionality. There is no consumer install-time execution, foreign AI-agent control-surface write, persistence, credential harvesting, or command-output exfiltration path established.
Decision evidence
public snapshot- package.json has only prepublishOnly; no install/postinstall hook for consumers.
- Entrypoints are SDK exports in dist/cjs/index.js and dist/esm/index.js, not import-time payload launchers.
- Cloud code in dist/esm/642.js requires explicit apiKey/CURSOR_API_KEY and user API calls for cloud agents.
- Runtime writes are local agent state files such as agents.ndjson, runs.ndjson, run_events.ndjson, checkpoints.ndjson under caller-selected/default state root.
- MCP servers, envVars, artifacts, and prompts are user-supplied SDK options, not harvested automatically.
- dist/esm/sqlite.js child_process/native-loading signals are bundled sqlite/local-runtime support, not lifecycle persistence or exfiltration.
Source & flagged code
6 flaggedA single source file combines environment access, network access, and code or shell execution; review context before blocking.
dist/esm/sqlite.jsView on unpkg · L1Findings
Affected versions and remediation
This report applies to @cursor/sdk@1.0.22.
See all versions of @cursor/sdk for other recorded verdicts.
- Review the evidence and your use of @cursor/sdk@1.0.22 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.