The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to read marketing context from .agents/product-marketing-context.md or .claude/product-marketing-context.md (found in SKILL.md). This creates a surface for indirect prompt injection where a malicious context file could attempt to influence agent behavior. However, this is considered low risk and a part of the skill's primary function.
[INGESTION_POINTS]: .agents/product-marketing-context.md and .claude/product-marketing-context.md.
[BOUNDARY_MARKERS]: Absent; the skill relies on direct reading of the context file.
[CAPABILITY_INVENTORY]: The skill does not possess any dangerous capabilities such as file modification, network egress, or shell command execution.
[SANITIZATION]: No sanitization or validation of external context content is mentioned or implemented.
[SAFE]: No obfuscation, data exfiltration patterns, or unauthorized command executions were detected. All tool integrations (Stripe, Rewardful, Tolt, etc.) are standard, well-known industry services.

referral-program