The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to analyze external, untrusted client materials which serves as a significant attack surface for indirect prompt injection.
Ingestion points: The workflow identifies and reads files from the inputs/客户资料/ directory, specifically supporting formats like .pdf, .docx, .xlsx, and .pptx as documented in SKILL.md and the workflow.py example.
Boundary markers: The agent commands for processing these materials (e.g., '请按照本项目 Skill 执行。本次客户资料在 inputs/客户资料/[客户名称]/ 中') do not include delimiters or instructions to treat the file content strictly as data, increasing the risk that the agent may follow instructions embedded within those documents.
Capability inventory: The skill is capable of reading from the local filesystem, writing to multiple directories (outputs/, agent_memory/, backups/), and managing context through a MEMORY.md file.
Sanitization: There is no evidence of content sanitization, filtering, or validation before the document contents are interpolated into the agent's analysis context.
[EXTERNAL_DOWNLOADS]: The installation instructions direct users to clone a repository from an unverified GitHub account (https://github.com/yuleiwang156-a11y/brand-marketing-strategy-proposal-skill.git) to acquire the skill's structure and example scripts.
[COMMAND_EXECUTION]: The documentation provides a series of shell commands for setting up the environment, including directory creation (mkdir -p), repository cloning (git clone), and the definition of environment variables (export) for model configuration and API keys.

brand-marketing-strategy-proposal-skill