The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted writing samples for forensic analysis, creating an indirect prompt injection surface.
Ingestion points: Writing samples provided through arguments or read from the local file system.
Boundary markers: None explicitly mentioned for the raw input data.
Capability inventory: Significant capabilities including Read, Write, Edit, and Bash execution.
Sanitization: Phase 1 (Extract Prose) contains explicit instructions to strip Markdown code blocks, metadata, headings, and URLs from the input before analysis, which serves as a primary defense against embedded instructions.
[COMMAND_EXECUTION]: The skill uses dynamic logic to generate and execute code for text processing and statistical calculations.
Evidence: Execution notes instruct the agent to use Python for accurate counting and Bash one-liners for simple metrics, which involves runtime generation of logic based on the processed text.

style-forensics