The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external blog URLs, creating a surface where malicious instructions embedded in those articles could influence the agent's analysis.
Ingestion points: External blog index and article URLs provided by the user at runtime (SKILL.md, Step 1 and 2).
Boundary markers: The skill lacks explicit delimiters or boundary markers to separate the untrusted blog text from the agent's processing instructions.
Capability inventory: The skill has the capability to execute shell commands (curl), write to the local file system (writing-styles/), and read local configuration files (AGENTS.md, CLAUDE.md).
Sanitization: No sanitization or filtering of the fetched content is specified before the agent performs its stylistic analysis.
Multi-step chain: The resulting style guide is intended to be used as context for future agent actions (rewriting articles), allowing potential injections in the source material to propagate and influence downstream tasks.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute curl commands to retrieve raw HTML content when standard web fetching tools are insufficient.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch data from arbitrary, user-supplied external URLs.

study-writing-style