The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of shell commands including gh repo create, curl, node, and npm to automate the site setup process.
[DATA_EXFILTRATION]: The skill reads authentication tokens from the local file system at ~/.aem/da-token.json and transmits them to vendor-controlled endpoints (e.g., admin.da.live and admin.hlx.page) to perform authorized actions.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs a token management utility directly from a vendor-owned GitHub repository using the command npm install -g github:adobe-rnd/da-auth-helper.
[DYNAMIC_EXECUTION]: JavaScript code snippets are executed via node -e to process local JSON configuration files and verify token expiration at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because user-provided variables are used in shell commands without validation.
Ingestion points: User-provided inputs for {{ORG}}, {{REPO}}, and {{SITE_NAME}} collected in Step 1.
Boundary markers: No delimiters or instructions are present to prevent the agent from interpreting malicious content within these variables.
Capability inventory: Uses subprocess calls for gh, curl, and npm across multiple steps in SKILL.md.
Sanitization: There is no evidence of input escaping or validation before these variables are passed to shell command templates.

create-site