The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external research files which could contain untrusted instructions. This data is used to 'refine' the draft outline, meaning malicious content in the research source could manipulate the agent's output.
Ingestion points: Reads research/conversations.jsonl and research/sources.json to refine the article outline.
Boundary markers: The skill does not define clear delimiters or specific instructions to ignore embedded commands within the research data.
Capability inventory: The skill has access to Write, Edit, and restricted Bash commands (mkdir, mktemp, date).
Sanitization: There is no evidence of sanitization or filtering applied to the content read from the research files before it is processed by the model.
[SAFE]: No other malicious patterns were detected. The skill uses restricted bash commands and lacks network access, limiting the potential impact of any injected instructions.

storm-outline