Security audit
Security checks for vulnerabilities and agentic risk
This skill is a plausible automated AI-briefing publisher, but it can write posts, use a local API key, run on a schedule, and push changes to GitHub with weak scoping and incomplete permission disclosure.
Install only if you intend this agent to publish to the specified GitHub-backed Jekyll site automatically. Review the referenced pipeline script first, use a narrowly scoped MiniMax key and GitHub credentials, confirm the launchd schedule, and consider adding a dry-run or approval step before any commit or push.
The skill declares only http.request and file.write permissions, yet documents auto-commit and push behavior, which implies additional capabilities such as invoking git, accessing repository credentials, and transmitting changes to a remote host. This creates a serious transparency and containment problem because the declared permission model understates what the skill may do.
The manifest-level description says the skill generates a weekday AI briefing, but the body also documents that it auto-commits and pushes changes to GitHub. That mismatch hides a material side effect from users and reviewers, which can lead to unanticipated repository modification and publication when the skill runs unattended.
The skill persists generated output by creating a Jekyll post and then auto-committing/pushing it, so its effects survive beyond the current session and may be publicly exposed. Persistent scheduled publication increases risk because errors, prompt injection from fetched content, or unwanted edits can accumulate or go live without human review.
- Fetches top 30 HN stories, filters for AI keywords
- Fetches top 10 AI-tagged dev.to articles
- Uses MiniMax to pick the 3 most interesting and write a punchy briefing
- Creates a Jekyll post in `_posts/YYYY-MM-DD-quick-hits-DAY.md`
- Auto-commits and pushes to GitHub
The skill states it uses MiniMax but does not clearly warn that content will be sent to an external service using a local API key. That omission matters because fetched articles or generated drafts may contain sensitive or unpublished material, and users are not given a clear privacy boundary up front.
The skill description does not prominently warn users that it writes files and automatically commits and pushes them to GitHub. Hidden persistent and publication side effects are dangerous because users may invoke or schedule the skill expecting only content generation, not repository mutation and remote distribution.
Automatically committing and pushing to GitHub goes beyond simple briefing generation and introduces a publication action that can change a live site or repository state. In a scheduled skill, this is especially risky because generated or malformed content can be propagated without an explicit user approval step.
The manifest frames the skill as fetching stories and generating a briefing, but the documentation indicates it reads a local secrets file to obtain an API key for MiniMax. Accessing local secret material is a sensitive capability that should be explicitly justified and scoped, especially since it is not mentioned in the manifest description or permissions.