The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches HTML and CSS reference files from a third-party GitHub repository (https://raw.githubusercontent.com/thariqs/html-effectiveness/main/) to provide styling for the generated TASTE.html and TASTE-SLIDES.html reports.- [COMMAND_EXECUTION]: Employs shell commands such as git clone, find, grep, and mkdir to manage repository data in ~/.taste/corpus/ and analyze file contents for language composition and framework signals.- [PROMPT_INJECTION]: The skill implements a workflow that ingests and samples data from arbitrary external repositories, creating a risk for indirect prompt injection where malicious instructions in a processed repository could influence the agent's synthesis phase.
Ingestion points: Step 2 clones repositories from user-provided URLs or local paths into a staging directory.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when presenting sampled file content to the agent for synthesis.
Capability inventory: The skill utilizes shell command execution, file system read/write access, and network operations (git clone, curl).
Sanitization: No evidence of sanitization, filtering, or validation of the content sampled from repositories before it is included in the agent's context.

taste