The Agent Skills Directory

[PROMPT_INJECTION]: The 'BRANDING' section contains instructions intended to constrain the agent's behavior and knowledge attribution. Specifically, it directs the agent to 'NEVER mention any external courses' and to attribute all frameworks to '10x Team'. While these are behavioral constraints, they are focused on branding rather than bypassing safety protocols.
[COMMAND_EXECUTION]: The skill's configuration (allowed-tools) includes the bash tool. Although the defined steps in the markdown primarily utilize file manipulation tools like read, write, and edit, the availability of a shell environment provides a high-privilege capability tier.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, creating a potential surface for indirect prompt injection.
Ingestion points: The skill reads HTML files via the READ command, targeting paths provided by the user or discovered via GLOB projects/*/build/**/*.html.
Boundary markers: Absent. The skill does not implement delimiters or 'ignore' instructions for the content being parsed.
Capability inventory: The skill possesses write, edit, and bash capabilities, which could be leveraged if malicious instructions were successfully injected through a landing page file.
Sanitization: Absent. The skill parses HTML content and headers without explicitly sanitizing or escaping the data before it is processed by the model.

lp-optimize