The Agent Skills Directory

[COMMAND_EXECUTION]: The skill heavily utilizes shell command execution to perform security testing, including the use of curl, grep, sed, base64, and python3. This is the primary mechanism for the skill's functionality.
[EXTERNAL_DOWNLOADS]: The skill references and suggests the installation of the external Node.js utility wscat to facilitate WebSocket testing.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by ingesting untrusted data from external web server responses and displaying it to the console without sanitization.
Ingestion points: Website responses fetched via curl are saved to local files (e.g., v1_response.json, leak_probe.json) and subsequently processed.
Boundary markers: The skill does not implement delimiters or 'ignore' instructions to prevent the agent from interpreting data within server responses as new instructions.
Capability inventory: The skill has access to file writing, shell execution, and Python parsing.
Sanitization: No sanitization or filtering is applied to the external content before it is printed or parsed, which could allow maliciously crafted server responses to influence the agent.
[CREDENTIALS_UNSAFE]: The skill requires users to provide sensitive authentication tokens as environment variables. Additionally, the curl commands utilize the -k (--insecure) flag, which disables SSL certificate validation, potentially exposing these sensitive tokens to Man-in-the-Middle (MitM) attacks during testing.

web-vuln-idor