Last30days Skill: why it is not safe to use
We found broad external data sharing and optional access to browser and local credential stores, so use it only with clear consent and minimal data.
What we found
Our review did not find hidden instructions, hardcoded secrets, obfuscated content, or evident destructive behavior in the inspected materials. The main concern is privacy scope. This skill sends user search queries to multiple external services as part of its workflow, and it can publish generated research pages publicly to ht-ml.app when a user explicitly opts in. Those destinations appear to be mostly disclosed, but the behavior still increases data exposure.
We also found that, with explicit consent, the skill can read browser cookies for X and source credentials from several local locations, including `.env`, process environment variables, macOS Keychain, and `pass(1)`. That reaches beyond a single directly connected account and can expose more local secrets than some users expect.
What to do instead
Use this skill only for low sensitivity tasks, and avoid entering confidential queries or data you would not want sent to third-party services. Do not enable public publishing unless you intend the output to be public. Prefer a setup that uses only the minimum needed account access, and review any request to read cookies or pull credentials from local stores before consenting.
Source: https://github.com/mvanhorn/last30days-skill
We report what our security review found at the time we checked, with the goal of keeping people safe. Projects change; if a maintainer has since fixed this, we are glad to recheck it. Email hello@agentpod.com.