The Agent Skills Directory

[SAFE]: The skill serves as a planning and educational framework. It does not contain any bundled binaries, obfuscated scripts, or persistent mechanisms. Its primary function is to guide an agent through authorized security testing phases using standard industry tools.
[INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect injection by instructing the agent to process data from external targets such as HTTP responses, JavaScript source code, and security tool outputs.
Ingestion points: Web application responses, JS files, and API data analyzed during Phases 1, 2, and 3 (SKILL.md).
Boundary markers: The methodology explicitly mandates the use of 'Marker Discipline,' requiring unique, random alphanumeric strings (8+ characters) to distinguish injected payloads from target content (references/details.md).
Capability inventory: The agent is instructed to use CLI tools like curl, nuclei, ffuf, and ghauri for testing (SKILL.md).
Sanitization: The 'Marker Discipline' section requires searching the baseline response for the marker string before claiming a reflection to prevent false positives and data confusion (references/details.md).
[DYNAMIC_EXECUTION]: The skill advises the agent to generate and execute local Python scripts when iterating over lists of more than 5 items to ensure proper error handling and avoid silent failures associated with shell loops (references/details.md).
[COMMAND_EXECUTION]: The methodology provides templates and workflows for executing common security reconnaissance and testing tools (e.g., subfinder, httpx, nmap). All such commands are subject to a 'Mandatory confirmation gate' requiring explicit user approval of the target, scope, and command effects (SKILL.md).

bb-methodology