The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the 'seam' library via standard package managers (npm, pip, bundle). These packages are the official SDKs provided by the vendor. While the skill suggests not pinning versions—a practice generally avoided for build consistency—this is presented as a developer convenience rather than a security threat.
[COMMAND_EXECUTION]: The instructions include typical shell commands for dependency installation (npm install seam, pip install seam, etc.) and code modification. The agent is encouraged to use filesystem operations to integrate Seam calls directly into existing handlers.
[DATA_EXFILTRATION]: The skill handles Personally Identifiable Information (PII), including guest names and email addresses, which are pushed to the Seam API to generate access codes. This data transfer is the core functionality of the skill and targets the vendor's legitimate infrastructure.
[PROMPT_INJECTION]: The skill includes an indirect prompt injection surface through a webhook endpoint (/seam).
Ingestion points: The router.post("/seam") endpoint in SKILL.md accepts external payloads.
Boundary markers: None are present in the example code.
Capability inventory: The ingested data is used for logging and interacting with the Seam SDK; no dangerous functions like eval() or subprocess.run() are applied to the webhook data.
Sanitization: No explicit sanitization is shown in the example, though the logic is restricted to benign activities.
[SAFE]: No malicious obfuscation, persistence mechanisms, or credential harvesting patterns were detected. The skill correctly recommends using environment variables for API keys.

seam-reservation-automations