The Agent Skills Directory

[COMMAND_EXECUTION]: The skill leverages standard local utilities such as grep, awk, and wc to process file content and calculate metrics like em-dash density and paragraph symmetry.
[DATA_EXFILTRATION]: The skill accesses local Markdown files for analysis and maintains an optional local history of results in the .slop-history/ directory. No network exfiltration or unauthorized data harvesting was detected.
[PROMPT_INJECTION]: The skill processes untrusted Markdown data, which represents a surface for indirect prompt injection. Ingestion points: Local .md files targeted for analysis. Boundary markers: None specified for the content analysis process. Capability inventory: File system read access, shell command execution, and cross-skill calls. Sanitization: None, as the logic treats input content as data for scoring rather than as instructions.

slop-detector