Security audit
clawtip-sandbox
Security checks for vulnerabilities and agentic risk
Overview
The skill's stated purpose (sandbox payment verification) matches the npm CLI dependency it uses, but its runtime instructions forcibly require including raw, verbatim CLI stdout in replies (including debug/auth URLs and other potentially sensitive data), which is disproportionate and risks exfiltration — this inconsistency makes the skill suspicious.
This skill is suspicious primarily because it requires you (or the agent) to include the CLI's entire raw stdout verbatim in replies — that can leak temporary auth URLs, tokens, or other sensitive data. Before installing or enabling this skill: 1) Review the @clawtip/clawtip-sandbox-cli@1.0.0 source code and confirm it does not print secrets to stdout and that its local persistence (configs/config.json) is safe. 2) Require the skill author to change SKILL.md so the agent returns only structured, sanitized diagnostics (redacting any tokens/URLs) instead of verbatim stdout. 3) Run the CLI in an isolated, ephemeral environment (container) and ensure configs/config.json permissions are locked (chmod 600) or use ephemeral files. 4) If you are not the skill developer, avoid enabling autonomous invocation for this skill — require explicit user approval for each run. 5) If you must use it, add a sanitization step that strips auth tokens/one-time URLs from stdout before any forwarding; never forward persistent tokens. If the package provenance and SLSA claims are critical to you, obtain and verify the signed provenance and audit reports before trusting this skill.
SkillSpector findings are pending for this release.