Security audit
Security checks for vulnerabilities and agentic risk
The skill is for Bear Blog post management, but its examples include unsafe authenticated browser automation that can publish, unpublish, and delete posts without clear user confirmation.
Install only if you are comfortable letting the agent operate an authenticated Bear Blog browser session. Review content, target post, slug, and visibility yourself before publishing or unpublishing, and do not use the documented confirmation-bypass deletion flow without a separate explicit confirmation and backup.
The documentation expands from blog post management into full account login automation, including credential entry and reliance on persistent session cookies. This increases the skill's authority and attack surface by enabling account access workflows that are broader than the stated purpose and could be misused to take over or operate a user's publishing account.
The skill documents arbitrary browser-side JavaScript evaluation to read and modify page state, which is more powerful than necessary for ordinary post editing. Exposing evaluate-based DOM/script execution makes it easier to perform unintended actions, scrape broader account data, or bypass normal tool safety constraints.
Overriding window.confirm to force deletion bypasses an explicit user-safety control built into the application. This makes destructive actions easier to trigger accidentally or programmatically and removes an important last-chance barrier against irreversible content loss.
The publish action is a state-changing operation that can make content public, but the documentation presents it as a routine click without any warning, confirmation, or review step. In an agent context, this raises the risk of unintended publication of drafts, sensitive text, or incomplete content.
The delete workflow not only performs a destructive action but also suppresses the site's confirmation mechanism, with no compensating safety guidance. In a browser automation skill, this materially increases the chance of irreversible deletion from a mistaken target selection or malicious prompting.
Unpublish changes public visibility and can affect availability, readers, and linked content, yet the documentation gives no warning that this is a user-visible state change. While less severe than deletion, it still creates avoidable operational risk in an automation context.
The workflow instructs the operator to enter account credentials directly into the browser tool and then proceed with authenticated actions, but it does not warn about sensitive credential handling, secret exposure in logs, or the risks of using persistent session cookies. In an agent setting, this can lead to accidental disclosure of credentials or unintended authenticated actions if outputs, traces, or session state are reused.
The publish step performs a real remote state-changing action by creating or publishing a live blog post, yet the workflow presents it as a routine step without an explicit confirmation or warning about irreversibility and production impact. In an agent-assisted workflow, this increases the chance of accidental publication, unintended content disclosure, or unauthorized modification of the user's site.
No suspicious patterns detected.
