The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill workflow involves an agent reading raw content from local article files (article.md and meta.json) to extract and summarize text for social media. This ingestion point lacks explicit boundary markers or sanitization instructions to prevent the agent from following instructions potentially embedded within those articles. Because the agent has access to powerful tools like Bash, Write, and Edit, this constitutes a vulnerability surface where a malicious article could attempt to hijack the agent's behavior.
[COMMAND_EXECUTION]: The skill heavily relies on the execution of local shell scripts (syndicate.sh, pick-next-article.sh, post-*.sh) to manage the syndication logic, state history, and API interactions. These scripts utilize common utilities like jq and curl, which are invoked via the agent's Bash tool throughout the process.
[EXTERNAL_DOWNLOADS]: The skill uses curl to interact with official social media platform APIs, specifically targeting bsky.social (Bluesky), api.linkedin.com (LinkedIn), and graph.threads.net (Threads). These network operations are used for the intended purpose of posting syndicated content to the user's accounts.

wjs-syndicating-articles