The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external content (news, blogs, user topics) to generate social media drafts, which introduces a potential surface for indirect prompt injection.
Ingestion points: Processes user-provided subjects, blog content, and news articles as described in SKILL.md.
Boundary markers: Implements a mandatory "Approval Gate" (3단계: 즉시 발행) where the agent must present the draft for user review and approval before proceeding.
Capability inventory: The skill uses threads_publish_text to write content to the Threads Graph API.
Sanitization: The human-in-the-loop requirement provides a manual validation step to ensure generated content is safe and intended.
[CREDENTIALS_UNSAFE]: The documentation provides guidance on configuring API credentials.
Evidence: SKILL.md contains instructions for setting THREADS_ACCESS_TOKEN and THREADS_USER_ID via environment variables using standard placeholders. This avoids hardcoding secrets and adheres to best practices for local development configuration.

threads-post-draft