The Agent Skills Directory

[Data Exposure & Exfiltration] (HIGH): The script scripts/text_diff.py allows reading and writing arbitrary files based on user-provided paths.
Evidence: The read_file function and the --output flag in main() use standard open() calls without any path sanitization or sandboxing, enabling access to sensitive files like ~/.ssh/id_rsa or .env.
[Indirect Prompt Injection] (HIGH): The skill possesses a high-risk surface for indirect prompt injection due to its core functionality.
Ingestion points: read_file in scripts/text_diff.py reads content from external, potentially attacker-controlled files.
Boundary markers: Absent; file content is directly processed and returned to the agent context.
Capability inventory: The skill has file-write capabilities via the --output flag in main().
Sanitization: Absent; the script does not filter or escape content from the files it compares.

text-diff