The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines and executes shell functions (rgw) using ripgrep or grep to audit local files. While these commands are scoped to the current directory and specific file extensions, they involve direct shell interaction with the host filesystem.
Evidence: Definition of rgw() function and multiple calls like rgw '✨|🪄|🌟' and rg -rln -i 'privacy' app/.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, such as website source code and live site content. There are no boundary markers or sanitization steps to prevent the agent from being influenced by malicious instructions embedded within the content being audited.
Ingestion points: Uses ripgrep/grep to read local codebase files and suggests fetching live site pages (homepage, pricing, product page).
Boundary markers: None present; the agent is instructed to read and interpret the content directly.
Capability inventory: File system read access via ripgrep/grep, network access to fetch external websites, and potential file write capabilities (implied by the instruction to 'fix only what is asked for').
Sanitization: No validation or filtering of the audited content is performed before processing.
[DYNAMIC_EXECUTION]: The skill dynamically generates and executes shell scripts by defining helper functions at runtime to perform its audit tasks.
Evidence: The skill instructs the agent to define the rgw() function in the shell and subsequently invoke it with various regex patterns.

fix-website