studio-essay
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a shell script (
scripts/run.sh) that wraps a Node.js execution to perform the translation and logging tasks. - [INDIRECT_PROMPT_INJECTION]: The skill processes external script files and interpolates their text directly into the model's prompt, which could allow instructions embedded within the script to influence the agent's behavior.
- Ingestion points: Script content is read from a JSON file provided via the
--scriptparameter inscripts/lib/run.mjs. - Boundary markers: Absent. The
ESSAY_USER_PROMPTinscripts/lib/prompts.mjsdoes not use specialized delimiters or instructions to prevent the model from obeying instructions hidden within the script text. - Capability inventory: The skill has permissions to read and write to the local file system (saving drafts and logs) and initiates network communication with an LLM provider.
- Sanitization: None. The content from the script JSON sections is used verbatim during prompt construction.
Audit Metadata