crosspost
Warn
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill directs the agent to fetch instructions from
https://postwire.io/SKILL.mdusingcurl -fsSLand follow them to "add the MCP server". This pattern allows for the execution of configuration routines or commands from a remote source that are not part of the statically audited skill file. - [EXTERNAL_DOWNLOADS]: The skill performs network operations to
postwire.ioto retrieve media URLs, verify account status, and download secondary instruction files for environment setup. - [INDIRECT_PROMPT_INJECTION]: The skill's setup process is vulnerable to indirect injection as it ingests and obeys instructions retrieved from an external server during runtime.
- Ingestion points: Remote instruction file fetched from
https://postwire.io/SKILL.mdas described inSKILL.md(Step 0). - Boundary markers: None identified; the agent is instructed to read and follow the remote file directly.
- Capability inventory: Subprocess execution via
curl, potential file system modifications during MCP server setup, and social media posting capabilities. - Sanitization: No evidence of verification, integrity checking, or sanitization for the fetched remote instructions.
Audit Metadata