The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The skill is designed to read sensitive API keys, application passwords, and secrets from ~/.claude/suparank-credentials.json and ~/.suparank/credentials.json. Accessing credential stores in the home directory is a high-risk operation.
[COMMAND_EXECUTION] (HIGH): The skill executes a Node.js one-liner using node -e to generate a JWT. It interpolates the ${admin_api_key} variable directly into a string literal within the command. If the configuration file contains a single quote (e.g., ' ; process.exit(1); //), an attacker could achieve arbitrary code execution on the user's machine.
[DATA_EXFILTRATION] (MEDIUM): The skill transmits credentials (via Authorization headers) to URLs defined in local configuration files (site_url, api_url, webhook_url). If these configuration files are manipulated to point to attacker-controlled domains, the credentials will be exfiltrated via the curl commands.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill processes untrusted article content from .claude/suparank-content/ to convert it to HTML and publish it.
Ingestion points: .claude/suparank-content/[folder]/article.md and metadata.json.
Boundary markers: Absent; the content is processed directly for payload construction.
Capability inventory: Shell execution (curl), Node.js execution (node -e), and file system access.
Sanitization: No evidence of sanitization or escaping for article content before inclusion in API payloads.

suparank/publish