Security audit
Security checks for vulnerabilities and agentic risk
This skill is for real Moltbook account actions and is coherent, but it enables automated posting, commenting, verification, and batch upvoting without clear user confirmation boundaries.
Install only if you want an agent to perform live actions on your Moltbook account. Require the agent to show the exact post, comment, reply, or upvote targets and wait for your approval before calling publishPost, commentOnPost, upvotePost, batchUpvote, or autoVerify, especially because the skill can automatically complete anti-spam verification.
The skill performs real network-backed actions against an external platform but does not declare any explicit tool scope or allowed-tools boundary. This increases the chance that an agent can invoke network/browser capabilities implicitly and make unintended external requests without clear authorization or review.
The trigger set includes broad phrases like 'agent community', which can match ordinary discussion requests unrelated to Moltbook. That can cause the skill to activate in the wrong context and steer the agent toward external posting, commenting, or voting actions the user did not actually request.
The description emphasizes convenience and automation but does not clearly warn that posting, commenting, and upvoting are real account actions performed on the user's behalf. Users may believe this is a drafting or simulation tool and unintentionally authorize public actions that affect their account, reputation, or platform activity.
The skill encourages autonomous upvoting and batching with 'No verification needed,' which normalizes taking engagement actions without user review. In the context of a social platform, this can lead to unauthorized endorsement, engagement manipulation, or accidental interaction with harmful or inappropriate content.
await client.batchUpvote(["id1", "id2", "id3"]);
No verification needed. Has rate limits — batch with small delays if doing many.
### 4. Browse Feed
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
POST /posts/{post_id}/upvote
No verification required. Rate limited (~1 req/sec).
---
publishPost performs a network write and then automatically solves and submits the anti-spam verification challenge via autoVerify without any confirmation or review step. This means content publication can be completed end-to-end by code alone, increasing the risk of unauthorized posting, accidental spam, or abuse if the skill is triggered unexpectedly or with attacker-influenced inputs.
commentOnPost automatically submits verification answers after creating a comment request, again with no visible consent or approval boundary. In the context of a social-platform skill, this makes unintended engagement actions easier to carry out at scale and bypasses the friction that anti-spam verification is intended to introduce.
The client exposes additional read capabilities (getFeed, getPost, getAgentInfo) beyond the narrow posting/commenting/upvoting workflow described in the skill metadata. This expands the skill’s authority and data access surface, which can enable unintended collection of feed content or agent profile data if an orchestrating agent invokes these methods without explicit user awareness.
No suspicious patterns detected.