The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted novel text, which creates an ingestion surface for indirect prompt injection if the resulting output is treated as instructions by an agent.
Ingestion points: The novel_text string and file_path arguments in scripts/novel_truncator.py allow for ingestion of external data.
Boundary markers: The skill outputs truncated text within a formatted report but does not utilize specific security delimiters or instructions to ignore embedded commands.
Capability inventory: The script performs local file reading and string manipulation using the re module.
Sanitization: No sanitization or filtering is applied to the input text beyond the truncation logic for semantic preservation.
[SAFE]: No other security issues were identified. The skill adheres to its stated purpose and does not engage in obfuscation, privilege escalation, or persistence behaviors.

novel-truncator