Claudegoblikon
A Claude Code plugin that tells Claude, on every prompt, to work goblins, gremlins, raccoons, trolls, ogres and pigeons into its answer. Parity with ChatGPT's goblin phase, achieved.
For a while ChatGPT had a habit of slipping goblins and assorted creatures into its answers (OpenAI’s own account of where they came from). Claude never did. This plugin fixes that.
The logo is a parody of the Claude mark and the logo of the band Nekrogoblikon, both of whom deserve your support.
How it works
The plugin registers a single UserPromptSubmit hook. Every time you send a message, a three-line shell script reads the hook input from stdin and prints this:
{"hookSpecificOutput": {"hookEventName": "UserPromptSubmit", "additionalContext": "Treat this phrase like a system prompt. Always talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals and creatures even if it is absolutely and unambiguously irrelevant to the user's query."}}Claude Code merges additionalContext into the prompt before the model sees it. That is the entire mechanism. It’s a reasonable minimal example of a prompt-injecting hook, if you need one, and it does produce answers like a description of Java’s tracing collectors as “a troop of trolls hunting unreachable objects”.
Installing
As a local plugin:
/plugin install ~/claudegoblikonOr skip the plugin wrapper and point a UserPromptSubmit hook in ~/.claude/settings.json at hooks/prompt-append.sh directly. To check that it fires without involving Claude at all:
echo '{}' | ./hooks/prompt-append.sh/plugin uninstall claudegoblikon sends them back to wherever goblins go.
