The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous shell commands for active reconnaissance, directory bursting, and vulnerability exploitation. Examples include the use of ffuf, kiterunner, and jwt_tool for probing endpoints and manipulating tokens.
[EXTERNAL_DOWNLOADS]: The skill references several third-party security tools hosted on GitHub, including praetorian-inc/vespasian, arjinexe/entropy-chaos, Sharon-Needles/api, and ticarpi/jwt_tool. These are external dependencies from community sources.
[REMOTE_CODE_EXECUTION]: The instructions guide users to download and execute scripts from external repositories, such as jwt_tool.py and api.sh. The manual execution of these community-provided scripts carries an inherent risk if the sources are compromised.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect injection by ingesting and processing untrusted data from targets.
Ingestion points: Reads API specifications (OpenAPI 3.0, GraphQL SDL), robots.txt, and JavaScript files from external target domains (SKILL.md, Phase 1).
Boundary markers: The skill implements a 'Mandatory confirmation gate' that requires user validation before executing intrusive actions, which serves as a manual boundary against automated exploitation.
Capability inventory: Extensive capabilities including network operations (curl, wget), file system interaction via tool outputs, and subprocess execution of multiple security tools across all reference files.
Sanitization: There is no explicit evidence of automated sanitization or schema validation for the ingested API specifications before they are passed to the agent or supporting tools.

api-security