The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Peppol network, including received invoices, credit notes, and inbound emails, which creates a surface for indirect prompt injection attacks.
Ingestion points: Data enters the agent's context through commands such as peppol inbox list, peppol mailbox get, and peppol document get --full as described in SKILL.md and various reference files.
Boundary markers: There are no instructions for the agent to use specific delimiters or to disregard instructions embedded within the processed invoice data.
Capability inventory: The agent can perform significant actions based on its analysis of this data, such as sending documents via peppol document send, creating new documents, or managing attachments.
Sanitization: The skill documentation does not outline any procedures for sanitizing or validating the content of the external data before it is used to drive agent logic.
[COMMAND_EXECUTION]: The skill facilitates the execution of the peppol CLI tool to interact with the file system and remote APIs.
Evidence: The skill uses tools to download attachments to local paths (e.g., peppol document attachment get -o output.pdf) and create backups of all document data to specified directories (peppol backup <dir>). These operations are consistent with the primary purpose of the skill as a document management tool.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the peppol CLI tool from external repositories.
Evidence: Installation instructions involve downloading code via Homebrew (brew install e-invoice-be/tap/peppol-cli) and Go (go install github.com/e-invoice-be/peppol-cli/cmd/peppol@latest). These sources are associated with the official vendor of the skill.

peppol