The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data including order notes, shipping addresses, and buyer names which could contain malicious instructions designed to manipulate the agent.
Ingestion points: External data is ingested through commands like orders get, orders +search, and orders fulfillments list as documented in SKILL.md and references/fulfillments.md.
Boundary markers: The skill does not implement explicit boundary markers or delimiters for untrusted data in the prompt instructions, though it does mandate re-confirming transaction details with the user.
Capability inventory: The skill possesses the capability to modify order status, delete records, and issue refunds using the shoplazza CLI.
Sanitization: No explicit sanitization or filtering of external data fields is described in the instruction files; the skill relies on the --dry-run safety protocol to prevent unauthorized execution.

shoplazza-orders