The Agent Skills Directory

[DYNAMIC_EXECUTION]: The prompt-engineer-toolkit component features a prompt_tester.py script that utilizes a --runner-cmd argument. This allows the AI agent to define a shell command for execution, incorporating content from prompt and input files via placeholder substitution. This architecture presents a high risk of arbitrary command injection if the underlying script does not rigorously sanitize the file content prior to shell execution.
[INDIRECT_PROMPT_INJECTION]: The skill suite includes multiple ingestion points for untrusted external data. For example, marketing-context scans the current codebase, site-architecture audits external sitemap.xml files, and x-twitter-growth retrieves competitor data through web searches. This creates an attack vector where malicious instructions embedded in these external sources could override agent behavior during processing.
[COMMAND_EXECUTION]: The skill relies on 27 specialized Python scripts stored within the skill directory to perform tasks like content scoring, competitor analysis, and pricing modeling. These scripts are executed locally to perform file operations and data analysis, representing a significant volume of executable logic shipped with the skill.

marketing-skills