The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use shell commands for project copying (cp) and batch-editing files. It explicitly highlights the risk of path injection by instructing the agent to quote all filesystem paths in shell commands.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from a local repository which could contain malicious instructions.
Ingestion points: The agent reads project source files, including HTML, CSS, JavaScript, README files, and manifests from a user-provided $SOURCE path.
Boundary markers: There are no specified delimiters or instructions to ignore instructions embedded within the source files being processed.
Capability inventory: The skill utilizes shell execution (batch edits, project copying, and temporary scripts) and browser automation (opening local Chrome/Chromium) to verify the output.
Sanitization: The skill mitigates shell injection by requiring path quoting and sanitizes derived folder names by normalizing them to lowercase kebab-case.
[DYNAMIC_EXECUTION]: The skill allows the creation and execution of temporary scripts for text replacement within the project and may perform JavaScript bundling during the runtime conversion phase.

extract-ds