The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were identified across the skill's instructions or reference documents.
[CREDENTIALS_UNSAFE]: The skill correctly handles authentication by identifying the need for API keys (pcs_live_*, pcs_test_*) and explicitly instructing users to store them in .env files rather than providing them in chat, which is a standard security best practice.
[DATA_EXFILTRATION]: Network operations are restricted to the vendor's official domain (portaly.ai). There are no patterns suggesting the exfiltration of sensitive local files or environment variables to unauthorized third parties.
[EXTERNAL_DOWNLOADS]: References to external tools are limited to the vendor's own ecosystem, such as the portaly-user skill from the portaly-ai organization. No unvetted or suspicious third-party dependencies are introduced.
[COMMAND_EXECUTION]: Shell command examples (e.g., curl) are used for legitimate API interactions. The skill does not instruct the agent to execute arbitrary or dangerous commands on the host system.
[PROMPT_INJECTION]: The instructions are focused on operational workflows and do not contain phrases designed to bypass safety filters, extract system prompts, or override the agent's core behavioral constraints.

portaly-email