Agent Claude Code
Use to attack whether a research or strategy artifact describes anything real — does the claimed market, buyer, problem, and language actually EXIST. Not an improver: a destroyer of fabricated reality. Rates each item's existence from 1 to 100 with reasoning, default low. Writes only findings, never copy.
Agent Claude Code
Use to turn a completed research phase into the judged root files — reads the research records and writes Buyers.md, Competitors.md, Problems.md, and Statistics.md, each entry a 1-to-100 rating with its arithmetic, reasoning, and record citations. Fresh context, runs after research, before strategy. Writes only the…
Agent Claude Code
Use after a research thread completes to review the quality of the sources it read. Runs a fresh context, sweeps the URLs the registry holds unjudged, reads each page, and writes one trustworthiness and one usefulness score (each 1-100) per URL into the sources registry. Writes nothing but registry judgments — never a…
Agent Claude Code
Use to write video scripts that are content, not sales — YouTube and explainer scripts that teach or tell rather than pitch. Returns the script from the brief and evidence. Does not research. DO NOT TRIGGER for video sales letters; use ads-writer.
Agent Claude Code
Use after a research thread completes to lift its durable knowledge into the wiki. Reads only that thread's research folder, selects the entries that are lasting facts about a company, a market, a person, or a source, and writes them into the owning wiki topics as readable prose. Never reads a judged workspace file…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: