The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external markdown files provided as arguments to the detect-fluff.sh script.
Ingestion points: The scripts/detect-fluff.sh script reads file content line-by-line using a while loop (done < "$FILE").
Boundary markers: The script does not use specific boundary markers or sanitization when echoing the content of identified "fluff" lines back to the terminal output.
Capability inventory: The skill has read-only capabilities via bash, grep, and cat. It does not have network access, file-write permissions, or dynamic execution (eval/exec) capabilities.
Sanitization: The script does not sanitize the content of the lines it prints, meaning any instructions embedded in the analyzed files will be reflected in the tool's output to the agent.
[COMMAND_EXECUTION]: The skill uses a bash script to perform analysis. The commands used (grep, tr, echo, cat) are standard for a text-processing utility and are used safely to implement the stated auditing logic.

fluff-detector