The Agent Skills Directory

[SAFE]: No security issues detected. The skill instructions focus on marketing strategy and adhere to established protocols.
[PROMPT_INJECTION]: The skill instructions do not contain any patterns attempting to override safety guidelines, bypass system constraints, or extract system prompts.
[DATA_EXPOSURE]: No sensitive file access (e.g., .ssh, .aws) or network-based exfiltration patterns were identified. Access to brand-related files like products.md and audience.md is within the intended functional scope for marketing assistance.
[REMOTE_CODE_EXECUTION]: No external package installations (npm, pip) or remote script executions (curl | bash) were found in the instructions or reference materials.
[OBFUSCATION]: No encoded content, zero-width characters, homoglyphs, or other forms of obfuscation were detected in the skill files.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not utilize the !command syntax in SKILL.md for executing shell commands at load time.
[INDIRECT_PROMPT_INJECTION]: Evaluated the surface for indirect injection via processed marketing data.
Ingestion points: Reads product and audience data from the brand-brain directory (e.g., products.md, audience.md).
Boundary markers: The skill uses a structured context matrix and startup protocols to manage information flow, though specific prompt delimiters are not explicitly defined in the logic.
Capability inventory: Limited to generating SMS marketing copy and strategies; no code execution, file system modification, or network capabilities are present.
Sanitization: Not explicitly implemented at the input level, but the limited capability tier (text generation only) renders this surface safe.

afa-sms