The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the dice-fm-pp-cli and dice-fm-pp-mcp binaries from the author's GitHub repository (github.com/mvanhorn/printing-press-library) and npm package (@mvanhorn/printing-press-library). These are vendor-provided resources for the skill's functionality.
[COMMAND_EXECUTION]: The skill executes shell commands using the installed dice-fm-pp-cli binary, passing user-supplied arguments through the $ARGUMENTS variable.
[COMMAND_EXECUTION]: The normalize command includes an optional --classifier-cmd flag that allows the agent to execute an arbitrary external command or script for data classification purposes.
[DATA_EXFILTRATION]: The skill features a --deliver webhook:<url> option that enables the output of commands—which can include sensitive event and fan data—to be POSTed to external HTTPS endpoints. While the skill documentation notes that this feature is hardened with SSRF guards and is blocked in certain contexts (MCP), it remains a primary vector for data transfer to external domains.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from the external DICE.FM API and local configuration files, creating a surface for indirect prompt injection.
Ingestion points: External data from DICE API (events, orders, fans, tickets), local feedback logs, and imported classification files (classified.json).
Boundary markers: No explicit boundary markers or instructions to ignore embedded content are defined for the processed data.
Capability inventory: The skill can execute shell commands (dice-fm-pp-cli), write files to the local system (--deliver file:), and perform network POST requests (--deliver webhook:).
Sanitization: The skill implements a pseudonymization mechanism that replaces fan identifiers (PII) with deterministic tokens by default, though this can be bypassed with the include_pii: true flag.

pp-dice-fm