The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from several sources including local context files (.agents/product-marketing-context.md), user-provided keyword exports (Ahrefs, SEMrush, GSC), and customer call transcripts.
Ingestion points: Instructions in SKILL.md guide the agent to read local project metadata and analyze uploaded CSV or text files containing customer feedback or SEO data.
Boundary markers: There are no explicit instructions to use delimiters or ignore instructions embedded within the processed data (e.g., within a customer transcript).
Capability inventory: The skill is restricted to information processing and strategy generation; it lacks dangerous capabilities such as network exfiltration tools, file-writing commands, or system-level execution.
Sanitization: No specific data sanitization or validation logic is implemented for the analyzed data.
[DATA_EXPOSURE]: The skill references reading a local file .agents/product-marketing-context.md. This is a legitimate use of local project context to ground the agent's strategy recommendations and does not involve unauthorized access to system-wide credentials or sensitive user directories.
[COMMAND_EXECUTION]: No shell command execution, subprocess spawning, or dynamic script generation was found. The skill operates entirely through natural language instructions and static reference documentation.

content-strategy