The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx -y @claude-flow/cli@latest to download an external utility from the npm registry at runtime.
[REMOTE_CODE_EXECUTION]: By using the @latest tag with npx, the skill executes the most recent version of an unverified third-party package, allowing for potential supply-chain attacks or arbitrary code execution during the memory storage process.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute shell commands, specifically for invoking the npx CLI and managing local storage operations.
[DYNAMIC_EXECUTION]: The skill employs browser_eval to execute JavaScript (document.cookie) within the context of external websites to extract session information.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect injection as it ingests data from untrusted web pages during the authentication flow.
Ingestion points: Web content and headers processed by browser tools during browser_open and authentication steps in SKILL.md.
Boundary markers: Absent; there are no explicit delimiters or instructions to the agent to ignore malicious content embedded in the login pages.
Capability inventory: Bash (for CLI execution), Write (for local file operations), and browser_eval (for script execution in page context).
Sanitization: The skill partially mitigates risk by using mcp__plugin_ruflo-core_ruflo__aidefence_scan to check cookies for PII/secrets before storage.

browser-login