The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to analyze untrusted text files and return analysis metrics along with fragments of the source text (e.g., in the contrast_matches and slop_word_hits fields). This behavior allows for indirect prompt injection if the analyzed content contains malicious instructions intended to influence the agent's next steps.
Ingestion points: The script scripts/slop-score/src/analyze.ts reads text data from user-provided file paths or standard input.
Boundary markers: Absent. The skill does not use delimiters or warnings to isolate the untrusted input text from the agent's instructions.
Capability inventory: The skill possesses file read/write capabilities (Bun.file, Bun.write) and stdout logging, but does not execute external commands or make network calls.
Sanitization: The input text is normalized for character encoding but is not sanitized to remove or escape potential prompt injection markers before being returned in the JSON output.

slop-score