The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill uses local project files to provide context, which introduces a surface for processing untrusted data.
Ingestion points: The skill instructions in SKILL.md direct the agent to read .claude/project-context.md or .cursor/project-context.md to identify product details and audience profiles.
Boundary markers: The skill does not implement delimiters or include instructions for the agent to ignore potentially malicious content within these external files.
Capability inventory: Analysis of SKILL.md reveals no dangerous capabilities; the skill does not use network tools, modify files, or execute shell commands.
Sanitization: There are no procedures described to sanitize or validate the content retrieved from the external context files.

content-strategy