write-like-meng-on-x
Pass
Audited by Gen Agent Trust Hub on Jul 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes several shell commands to manage its data and validate its state. This includes running
git status,git diff, andgit commitfor version control. It also usesrg(ripgrep) for searching the tweet corpus and local documentation. - [COMMAND_EXECUTION]: The skill runs a bundled Node.js script (
scripts/update-tweet-corpus.mjs) to ingest new data and a Python validation script (quick_validate.py) located in the workspace. These tools are used for data maintenance and quality assurance. - [EXTERNAL_DOWNLOADS]: The skill uses an in-app browser to access
x.comto collect recent posts from the author's profile. This data collection is necessary for the 'Refresh' functionality described in the instructions. - [SAFE]: Indirect Prompt Injection Surface: The skill ingests data from external X posts into its local corpus (
references/tweet-corpus.jsonl). - Ingestion points: External data is collected from
x.comand stored inreferences/tweet-corpus.jsonlvia a batch update process. - Boundary markers: The instructions explicitly direct the agent to treat the corpus as 'voice evidence' rather than direct text to splice, and to exclude 'quoted-source text' from the voice field.
- Capability inventory: The agent has the ability to execute shell commands (
git,node,python3,rg), write to the local file system via the update script, and use an integrated browser. - Sanitization: The ingestion script (
update-tweet-corpus.mjs) normalizes and hashes the text for deduplication but does not perform content-based sanitization for potential prompt injections.
Audit Metadata