The Agent Skills Directory

[DATA_EXFILTRATION]: The scripts/text_truncator.py file contains a truncate_text_from_file method that reads files from arbitrary paths provided as arguments using the open(file_path, 'r') command. This enables an agent to access and expose sensitive local files, such as SSH keys, AWS credentials, or configuration files, to its context.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text input and file contents without sanitization or clear boundary markers, making it a surface for indirect prompt injection attacks.
Ingestion points: Untrusted data enters the agent context via the text argument in truncate_text and the content of files read via truncate_text_from_file (scripts/text_truncator.py).
Boundary markers: None identified in the prompt instructions or code to distinguish between system instructions and processed content.
Capability inventory: File system read access through the truncate_text_from_file function in scripts/text_truncator.py.
Sanitization: No filtering or validation is performed on the input text before processing and outputting it.

text-truncator