The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to interact with the environment. Specifically, it executes curl to retrieve server headers from a target URL and cat to read the contents of files located in recon/$TARGET/live-hosts.txt. While these are functional requirements for a penetration testing orchestrator, they involve executing system commands based on user-provided target variables.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data that could be controlled by an adversary to influence the agent's behavior.
Ingestion points: Data enters the context via the output of curl -sI "https://$TARGET" (network headers) and the content of the local file recon/$TARGET/live-hosts.txt.
Boundary markers: No explicit delimiters or instructions are used to tell the agent to ignore potentially malicious content within these data sources.
Capability inventory: The skill has the capability to execute shell commands (curl, cat, test, tr) and dynamically load a wide array of high-privilege attack skills (e.g., hunt-rce, okta-attack, m365-entra-attack) based on the parsed data.
Sanitization: The skill performs minimal sanitization (removing carriage returns via tr -d '\r'), which does not protect against malicious instructions embedded in the headers or the reconnaissance files.

hunt-dispatch