The Agent Skills Directory

[PROMPT_INJECTION]: No direct instructions to bypass safety guidelines or override behavior were detected within the skill documentation.\n- [PROMPT_INJECTION]: The skill ingests untrusted text for poll questions and options, presenting an indirect prompt injection surface. (Ingestion points: 'question' and 'options' parameters; Boundary markers: None; Capability inventory: Executes local Node.js scripts and performs network operations to chat APIs; Sanitization: Not documented).\n- [COMMAND_EXECUTION]: The skill calls a local script ('poll.js') to perform its functions, which is standard procedure for tool-based skills.\n- [DATA_EXFILTRATION]: Credential handling is performed via environment variables as required, and there is no evidence of unauthorized data transmission to unknown domains.\n- [SAFE]: All behaviors and resources are consistent with the skill's stated purpose of facilitating team decisions via polls on well-known platforms.

poll-create