The Agent Skills Directory

[SAFE]: The skill package is composed strictly of documentation files (SKILL.md, references/copy-frameworks.md, references/natural-transitions.md). No scripts, binaries, or system-level configuration files are included.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an ingestion surface for external data by instructing the agent to read context from .agents/product-marketing-context.md and user input. The evidence chain is as follows: 1. Ingestion points: local context files and user prompts; 2. Boundary markers: none present; 3. Capability inventory: text generation only, with no subprocess, network, or file-write tools; 4. Sanitization: none present. Given the lack of dangerous capabilities, this surface is considered low risk and inherent to the skill's purpose.
[SAFE]: No evidence of prompt injection, obfuscation, data exfiltration, or persistence mechanisms was found. The skill operates within a restricted text-generation scope.

copywriting