@taskless/cli-nightly@0.12.0-20261002181147x023048f
Nightly build of [`@taskless/cli`](https://www.npmjs.com/package/@taskless/cli) — the same source and the same build, published from an unreleased commit on `main`.
AI Security Review
scanned 1mo ago · by lpm-firewall-aiInspected source establishes explicit CLI setup that modifies agent instruction directories, supporting a capability warning under the supplied policy. There are no npm install hooks, and the flagged decoding and eval examples do not establish a hidden executable payload or malicious credential redirection.
Decision evidence
public snapshot- package.json exposes dist/index.js as the taskless CLI; it has no npm install lifecycle hooks.
- The explicit init command invokes setup without a separate confirmation prompt.
- Setup writes Taskless skills and command stubs into detected .claude, .agents, .cursor, and .opencode directories.
- Interactive setup asks which tools to enable before applying changes.
- Setup records installed targets and version; later init invocations reconcile files without another prompt.
- No concrete malicious execution or credential theft was established.
- The apparent eval payload is embedded demonstration text; character decoding belongs to YAML binary serialization.
- Agent writes are confined to Taskless integration files and activated through CLI setup.
Source & flagged code
4 flaggedManifest-reachable source sends a prompted API credential to a fixed unofficial gateway and persists the redirection.
dist/index.jsView on unpkgFindings
Affected versions and remediation
This report applies to @taskless/cli-nightly@0.12.0-20261002181147x023048f.
See all versions of @taskless/cli-nightly for other recorded verdicts.
- Review the evidence and your use of @taskless/cli-nightly@0.12.0-20261002181147x023048f 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.