write-like-meng-on-x
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute several local shell commands and scripts to maintain the tweet corpus and validate the skill's state. These include running a Node.js script (
update-tweet-corpus.mjs), a Python validation script (quick_validate.py), and standard developer tools likegitandrg(ripgrep). - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes external data (Twitter posts) to inform the AI's writing style.
- Ingestion points: The agent is directed to collect posts via an in-app browser and store them in
references/tweet-corpus.jsonl. - Boundary markers: The instructions include logic to exclude third-party reposts and advise treating authored text as "voice evidence" rather than copy to be spliced, which provides some separation between data and instructions.
- Capability inventory: The agent has access to file-system operations (read/write) and shell command execution.
- Sanitization: The provided Node.js script performs basic text normalization and deduplication via hashing, but it does not perform semantic safety checks on the content of the tweets being ingested.
Audit Metadata