The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the pbcopy system command to transfer generated content directly to the user's clipboard for easy pasting into editors.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external content from social media platforms like 'X' (formerly Twitter). This constitutes an attack surface where untrusted data enters the agent context.
Ingestion points: Phase 0 processes external 'X' posts, topics, or fragments provided by the user.
Boundary markers: The skill requires the agent to summarize the core claim and wait for user confirmation (Phase 0) before proceeding, serving as a manual checkpoint.
Capability inventory: The skill has the ability to write files to the vault (_草稿/公众号短文/), execute the pbcopy command, and generate images.
Sanitization: The skill performs extensive content-level sanitization based on domestic platform 'red lines' (forbidden words, model compliance), though it does not explicitly mention technical escaping of prompt characters.
[SAFE_PRACTICE]: The skill contains explicit security instructions regarding credential management, mandating that image generation API keys remain in environment variables and are never written to disk.

aikey-post