The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides several complex shell scripts for the agent to execute, involving loops, conditional logic, and pipes. It utilizes standard security tools such as curl, ffuf, hydra, and nuclei. While these are legitimate for security auditing, the broad use of shell commands creates a large capability surface for the agent.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze data from external target endpoints (e.g., login and OTP verification responses). This untrusted data is processed via shell tools like diff, grep, and awk without explicit boundary markers, posing a surface for indirect prompt injection where malicious server responses could attempt to influence the agent's logic.
Ingestion points: Remote server response bodies fetched via curl and stored in temporary files (e.g., /tmp/bf_body, /tmp/bf_l) as seen in Phase 1, Phase 2, and Phase 3.
Boundary markers: None observed; the agent is instructed to directly process the output of network requests.
Capability inventory: The skill has access to network operations (curl, ffuf, hydra), file system writes (/tmp/), and general shell execution.
Sanitization: No sanitization or validation of the remote response content is performed before processing.
[DYNAMIC_EXECUTION]: The skill employs python3 -c to execute small code snippets for entropy calculations and to generate dynamic inputs for ReDoS (Regular Expression Denial of Service) testing. This represents the dynamic generation and execution of code at runtime, although it is used here for benign mathematical and logic tasks.

hunt-brute-force