The Agent Skills Directory

[PROMPT_INJECTION]: The skill utilizes a local file references/learnings.md to store and retrieve persistent knowledge. This behavior creates a surface for indirect prompt injection where untrusted data or instructions encountered during a session could be saved and subsequently influence the agent's behavior in later invocations.
Ingestion points: The agent is instructed to read references/learnings.md at the beginning of Step 1 to gather context.
Boundary markers: There are no instructions to the agent to treat the content of the learnings file as untrusted or to ignore any embedded commands within it.
Capability inventory: The skill performs file read and write operations on its own reference files. No broader system-level execution or unauthorized network access was detected.
Sanitization: The skill does not provide any logic for validating or sanitizing user-provided or observed information before appending it to the learnings file.

sales-launchday