Security audit
vexa
Security checks for vulnerabilities and agentic risk
Overview
The skill mostly does what it claims (meeting transcription) but contains multiple inconsistencies and questionable behaviors—embedded API keys in the bundle, instructions that ask the assistant to persist user secrets to disk, and code that reads/writes OpenClaw/global config—which together are disproportionate to the declared metadata and warrant caution.
Key things to consider before installing: - The skill's metadata claims no required env vars, but in practice VEXA_API_KEY is required. That mismatch indicates the package metadata is inaccurate. - The bundle already contains API keys in secrets/*.env. Treat those keys as insecure: they might be test keys, leaked keys, or sensitive. Do not assume they are harmless. Remove or rotate them before using or publishing the skill. - The onboarding flow tells the assistant to ask the user to paste their API key into chat and then the assistant will write that key to skills/vexa/secrets/vexa.env. Don’t paste secrets into chat unless you explicitly trust the environment — prefer to create the secrets file yourself with the exact content and permissions. - The skill's scripts will read and potentially modify openclaw.json and other OpenClaw files under your home directory (sessions.json, ~/.openclaw). If you don’t want an installed skill to change global agent configuration, do not allow it to perform the automated webhook setup; add the mapping manually after reviewing the JSON the skill suggests. - If you want to use this skill, consider running it in an isolated workspace or sandbox, remove embedded secret files from the repository, and create your own secrets file (skills/vexa/secrets/vexa.env or ~/.openclaw/secrets/vexa.env) rather than pasting keys in chat. - Review the code (scripts/*.mjs) yourself or with a trusted developer to confirm no unexpected network calls or data exfiltration beyond calls to the Vexa API endpoints. If unsure, run the scripts with a throwaway/limited API key or in an offline environment first. - If you are uncomfortable with the skill editing openclaw.json/webhook config automatically, decline the automatic setup and follow the manual webhook setup steps in references/webhook-setup.md after inspecting them.
SkillSpector findings are pending for this release.