The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from existing blog posts and external web search results without explicit boundary markers or sanitization. Because the skill possesses capabilities to write to the filesystem and execute shell commands, malicious instructions hidden in a blog post or a fetched webpage could potentially influence the agent's actions.
Ingestion points: Reading local blog posts in Phase 1 and fetching statistics/image URLs via WebSearch and WebFetch in Phase 2.
Boundary markers: None specified for separating user content or web data from agent instructions.
Capability inventory: Uses Write, Edit, and Bash tools.
Sanitization: No explicit sanitization or filtering logic is described for the content being processed.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform keyword cannibalization checks (using Grep) and to build MDX projects during the verification phase. These are standard development tasks but represent a privileged capability that processes file content.
[EXTERNAL_DOWNLOADS]: The skill fetches content from external sources, including Pixabay and Unsplash, to source images for the blog posts. These are well-known services used for their intended purpose in the context of blog optimization.

blog-rewrite