SuperAPI AI Directory

wjs-voicedrop-post-processing

Writing & research

View repository

Updated 10-15 23:32

About this skill

VoiceDrop article post-processing

Input: article stem, provided by the calling parameters ($ARGUMENTS).

Runtime environment: unattended claude -p, the voicedrop MCP is already connected (read article/write version tools are available; publish, delete, community, and money-spending tools have been disabled by the caller). Do not ask any questions; exit as soon as it finishes.

Process

  1. Use the voicedrop MCP's read article tool to retrieve the full text and metadata of the article for that stem. If it cannot be read, retry once; if it still fails, exit with a non-zero status (the poller will retry in the next round).
  2. [Post-processing action — to be defined by Jianshuo, currently a skeleton version: only check, do not modify the article]
    • Check that the body is non-empty and the title exists
    • Finally output one line:postprocess ok: <stem> — <标题>(<字数>字)

Iron rules (same as the night-shift editor)

  • Never publish, never delete, never touch the style library or the community.
  • If the article is to be modified in the future, it may only be done via "write a new version", without switching head.

To be defined

Once you have decided what the post-processing specifically does (image matching / tagging / quality scoring / summary generation...), just replace step 2; the poller does not need to change.