The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines shell scripts that use curl, grep, and shasum to perform target fingerprinting and baseline checks.
The scripts rely on external variables such as $TARGET and $HOSTS, and read from local files like live-hosts.txt to construct and run commands.
[PROMPT_INJECTION]: The skill ingests data from external targets via curl and parses the output to determine which platform-specific attack skills to load.
Fingerprinting logic scans target response bodies for specific framework markers such as __NEXT_DATA__ or laravel_session.
This represents an indirect prompt injection surface where a malicious target could attempt to influence the agent's behavior by returning crafted strings.
[CREDENTIALS_UNSAFE]: The skill instructions provide workflows for handling sensitive authentication data, such as session cookies and bearer tokens, for greybox testing.
Detailed steps are provided for using $SESSION_COOKIE and $TOKEN variables in curl preflight probes.
The skill includes defensive instructions to avoid logging or persisting these credentials.

hunt-dispatch