The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references an external repository for the 'OktaTerrify' tool (github.com/silverhack/OktaTerrify). The source organization is not categorized as a trusted vendor, posing a risk for users who might download and execute unverified third-party security tooling.
[DATA_EXFILTRATION]: The skill uses 'attacker.example.com' as a destination URL in OIDC redirect tampering tests. This domain has been flagged by automated scanners for association with phishing activities. Using known malicious or high-risk domains, even for testing purposes, can lead to unintended data exfiltration or interaction with malicious infrastructure.
[COMMAND_EXECUTION]: The skill makes extensive use of powerful command-line utilities including curl for network requests and dig for DNS reconnaissance. These tools are used to perform tenant discovery, user enumeration, and interaction with Okta APIs.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to how it handles untrusted data.
Ingestion points: The skill fetches data from external Okta API endpoints (e.g., /api/v1/authn, /.well-known/openid-configuration, /api/v1/users) and pipes the output into shell tools and Python formatters.
Boundary markers: The instructions do not define boundary markers or delimiters when interpolating external API responses into the agent's context.
Capability inventory: The skill possesses network capabilities (curl, dig) and local execution capabilities (python3).
Sanitization: There is no evidence of sanitization or validation of the JSON structures returned by external APIs before they are processed by the agent.

okta-attack