The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes shell commands via the wewrite CLI tool to perform learning tasks. It interpolates local file paths such as {draft_path} and {final_path} directly into bash commands, which could lead to command injection if paths are not properly sanitized.
[EXTERNAL_DOWNLOADS]: The skill uses WebFetch and CLI tools to download content from external URLs (e.g., wewrite fetch-article <url>) to extract layout themes and import exemplars into the local database.
[REMOTE_CODE_EXECUTION]: The instructions guide the agent to install the wewrite CLI tool using uv tool install wewrite or by executing a local install.sh script if the command is not found in the environment. This involves downloading and executing code from external sources to enable skill functionality.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external URLs and articles.
Ingestion points: External articles are fetched via wewrite fetch-article <url> and saved to /tmp/article.md for processing.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when reading the content of these external files.
Capability inventory: The agent has access to Bash, Write, Edit, and WebFetch tools, allowing it to modify files and execute system commands.
Sanitization: There is no mention of sanitizing or filtering the content fetched from the web before it is analyzed or used to generate rules.

wewrite-learn