The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted external sources, such as web content crawled by cewler and search engine results harvested by theHarvester. This creates a vulnerability surface where malicious instructions or adversarial content embedded in a target's website or public records could influence the agent's context.
Ingestion points: Target website content (Stage 1) and public search engine/OSINT data (Stage 3).
Boundary markers: None specified in the instructions to delimit external content from system instructions.
Capability inventory: The skill utilizes network operations for authentication attempts (urllib, trevorspray) and writes audit logs to the local file system.
Sanitization: No explicit sanitization or filtering of harvested text is documented before it is processed by the wordlist engine or OSINT pipeline.
[COMMAND_EXECUTION]: The instructions describe a pipeline that relies on executing several external command-line tools, including cewler, theHarvester, hashcat, CrossLinked, and trevorspray. These are managed through referenced shell scripts (tools/osint_employees.sh, tools/wordlist_engine.sh) and pure Python modules (_spray_http_form.py). While these tools are standard for security research, the skill enables their automated orchestration against remote hosts.

credential-attack