The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines an ingestion point for external data by instructing the agent to read the local file .claude/product-marketing-context.md (SKILL.md). While explicit boundary markers and sanitization logic are absent in the instructions, the skill does not possess any risky capabilities (such as shell command execution, file-writing, or network operations) across its components (SKILL.md, affiliate-programs.md, program-examples.md), which significantly limits the impact of potential indirect prompt injection attacks.
[SAFE]: The skill consists exclusively of informational markdown content and business strategy guidance. It does not include any scripts, executable binaries, or automated network requests to external domains.

referral-program