The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data sources including file contents (via Bash(cat:*)), tool outputs (such as system-health or adoption-report), and user-supplied data to populate a pitch template. This creates an attack surface where malicious instructions embedded in these external sources could influence the agent's behavior.
Ingestion points: Project files read via cat, system-health outputs, and adoption-report outputs mentioned in Step 1.
Boundary markers: The skill lacks explicit instructions for the agent to use delimiters or sanitization when interpolating data from these files into its output.
Capability inventory: The skill utilizes Write to generate documents and Bash(cat:*) to read local file contents.
Sanitization: No sanitization or escaping of external content is specified before interpolation into the prompt context.
[EXTERNAL_DOWNLOADS]: The instructions direct the user toward an external script (verify-install.sh) and a third-party installation command (npx skills install) to manage dependencies and verify the environment if the local installation is incomplete.

system-pitch