The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script located at .claude/tools/scaffold_workshop.py using the uv tool. It interpolates user-provided arguments (directory name, title, and description) into a shell command. While these are wrapped in double quotes, there is a theoretical risk of command injection if the underlying shell environment incorrectly handles specific metacharacters inside the input strings.
[EXTERNAL_DOWNLOADS]: The skill uses WebSearch and WebFetch to retrieve information from external sources, specifically targeting AWS documentation and SDK references. While the intended targets are trusted/well-known services, these tools can technically access any URL returned by a search engine.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests untrusted data from the web and uses it to generate executable code (clean_resources.py) and documentation.
Ingestion points: Data returned by WebSearch and WebFetch tools (SKILL.md, Step 3).
Boundary markers: None explicitly implemented to isolate web-fetched content from the agent's instructions.
Capability inventory: The skill possesses Write, Edit, and Bash tools, allowing it to commit the drafted content and code to the local filesystem.
Sanitization: No sanitization or validation is performed on the content retrieved from the web before it is used to replace TODO markers in templates.

scaffold-workshop