The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and analyze untrusted data from external web servers.
Ingestion points: Target server responses (HTML bodies and headers) are fetched via requests and curl for analysis.
Boundary markers: There are no specific delimiters or instruction-ignore markers wrapping the content processed from external targets.
Capability inventory: The skill possesses network capabilities (via curl and Python requests) and the ability to process and report on extracted data.
Sanitization: Data extracted from target responses (such as ViewState values) is handled via regular expressions but is not sanitized against adversarial instructions before being presented to the agent.
[REMOTE_CODE_EXECUTION]: Provides a diagnostic Python script that utilizes the requests library to test server-side parsing logic. The script does not execute remote code locally or perform unsafe operations like eval() on data received from the network.
[COMMAND_EXECUTION]: Contains example curl commands for fingerprinting server configurations and checking for the presence of diagnostic endpoints like trace.axd. These are standard security auditing procedures.
[DATA_EXFILTRATION]: Network activity is restricted to probing user-defined targets for vulnerability research. No logic was found that attempts to exfiltrate local secrets, credentials, or system information to unauthorized third parties.

hunt-aspnet