The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses eval "$(stealth_curl_headers)" in SKILL.md to execute shell commands generated at runtime by a function. This function is defined in a sourced script (stealth-profile.sh) located in a directory ($SUPERHACKERS_ROOT) that the skill attempts to auto-detect in the user's home directory. This allows for the execution of non-static code that could be subverted if the environment is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill processes raw response data from external, untrusted network targets via curl and vulnerability scanners. It lacks boundary markers to isolate target data from agent instructions.
Ingestion points: Raw HTTP response bodies and headers captured by curl, and output files from nuclei and nikto (referenced in SKILL.md).
Boundary markers: Absent. The instructions do not direct the agent to treat the captured network data as untrusted or to ignore instructions embedded within it.
Capability inventory: Execution of shell commands, network request generation, and file system interactions.
Sanitization: None. The captured data is processed directly by terminal utilities and the agent without filtering.
[COMMAND_EXECUTION]: The skill relies extensively on shell command execution for its primary operations. This includes a custom perl fallback for the timeout command in SKILL.md that uses exec to run arbitrary arguments provided to it, and the execution of scripts within the $SUPERHACKERS_ROOT framework.
[EXTERNAL_DOWNLOADS]: The documentation recommends the installation of third-party security tools from remote repositories and package registries, such as nuclei from GitHub and sqlmap from PyPI. While these are recognized security tools, they represent unverified external dependencies.

vulnerability-verification