The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied JSON payloads containing market research and reviews. These inputs are ingested through the apply and add-review commands in scripts/arsenal.py. The skill generates HTML and Markdown reports (file-write capability) that incorporate this data. To prevent injection or breakout attacks, the skill implements boundary markers and sanitization via html.escape for HTML templates and a dynamic backtick-fencing mechanism in scripts/render.py to ensure JSON blocks in Markdown cannot be terminated by malicious content.
[DYNAMIC_EXECUTION]: The skill uses __import__ to dynamically reference a local module and importlib within its test suite. These operations are restricted to hardcoded local file paths and do not facilitate arbitrary code execution from untrusted sources.
[COMMAND_EXECUTION]: The skill provides a CLI interface for workshop management. All script operations are performed on local files within the workspace and do not invoke external shell commands using unsanitized user input.

growth-arsenal-workspace