The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill collects detailed sales information from the user and writes it to a persistent file (.agents/sales-context.md). This file is intended to be read by all other sales-related skills to provide context. If a user provides malicious instructions within their answers (e.g., in the value proposition or company description), these instructions could influence the behavior of other agents that consume this file.
Ingestion points: User responses to conversational questions in SKILL.md and the existing .agents/sales-context.md file.
Boundary markers: The skill uses Markdown headers and bullet points to structure the data, but it lacks specific instructions to the agent to treat the ingested data as untrusted or to ignore embedded instructions.
Capability inventory: The skill has the ability to read and write to the filesystem (specifically the .agents/ directory). It does not have network access or the ability to execute shell commands.
Sanitization: No input sanitization or validation is specified before the user input is written to the output file.

sales-context