The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill instructs the installation of an external package using pip3 install nosqlmap in Phase 5, which introduces code from an unverified source.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill initiates network connections to external target systems using curl to test for vulnerabilities and extract data, which could be leveraged for data exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes responses from untrusted external targets without sufficient sanitization or security boundaries.
Ingestion points: External content is ingested via curl responses stored in the RESP variable during character enumeration (Phase 4) and via response timing analysis in Phase 3 and Phase 6.
Boundary markers: The skill does not implement any boundary markers or instructions to the agent to ignore potentially malicious instructions embedded in target responses.
Capability inventory: The skill has the capability to execute shell commands (curl, pip3, echo, wc) and perform network operations.
Sanitization: There is no evidence of sanitization or validation of the content received from external targets before it is processed or printed to the console.

hunt-nosqli