Security audit
Security checks across malware telemetry and agentic risk
This is a disclosed Postproxy helper for managing real social media accounts, with powerful posting, deletion, messaging, queue, and webhook features that are expected for its purpose.
Install only if you trust Postproxy with the connected social accounts and API key. Confirm the exact content, target profiles/placements, DM recipient, delete scope, queue impact, and webhook destination before acting; prefer drafts or pausing queues when unsure, and keep API keys and webhook secrets out of logs and chat output.
This skill publishes to, modifies, and deletes content on the user's real social media accounts, and can read/send private messages. Before any outward-facing or irreversible action, summarize exactly what will happen and get explicit user confirmation:
- **Publishing** (create post without `draft`, publish draft, comment, reply): content goes live publicly and immediately. State the exact text, media, and target profiles/placements first. Prefer creating a draft when intent is ambiguous.
- **Deleting**: deleting a post from the DB only (`DELETE /api/posts/{id}`) leaves it live on the platforms; `delete_on_platform=true` or `/delete_on_platform` removes it from the social networks **irreversibly**. Never delete on platform without the user explicitly confirming the post and scope.
- **Direct messages and private replies**: these are private communications with real people. Only read or send them when the user explicitly asks; don't quote DM contents into other posts, tools, or external services. Private replies bypass Meta's 24h window — confirm before initiating contact.
- **Placements**: when posting to Facebook without an explicit placement, the post goes to a **random connected page**; LinkedIn defaults to the personal profile. Always resolve and confirm the placement for placement networks.
- **Webhook secrets**: the `secret` returned when creating a webhook is a credential — never print it in output, log it, or commit it. Webhook payloads can contain private content (DMs, comments) and are sent to the configured URL — only register HTTPS endpoints the user controls and trusts.65/65 vendors flagged this skill as clean.
No suspicious patterns detected.