saatvik333

GitHub profile1 plugin in the catalog

Plugins by saatvik333

1 plugins by saatvik333 in the catalog, most-starred first.

saatvik333skill-courtPuts every skill your agent writes on trial: replays real tasks with and without the skill in paired Docker sandboxes, scores each pair with the task's check or a blind pairwise judge, and gives a verdict per model. A skill that hurts is hidden from that model only (skill index, skills_list, skill_view), with the cause and the SKILL.md line behind it; an appeal drafts a fix, proves it on the same tasks and writes it only when you apply it (undo restores the original). Optional automatic trials after the agent creates or patches a skill. CLI (hermes court trial/report/appeal/auto/tasks/quarantine/release/list) plus a Desktop page. Disclosure: trials run hermes and docker/podman commands, create and delete two replay profiles cloned from yours (config, API keys, SOUL, memory, skills; never messaging channels) and spend model tokens with your provider (per-trial budget, default $3; automatic trials: daily $2 and 300,000 judge tokens). Replays are single-query sessions with approvals denied, never -z. The judge (your own provider, via ctx.llm) sees task text, both runs' answers and sandbox files, and for a cause a transcript excerpt and the skill. Automatic trials start a background `hermes court trial --auto`; verdicts are posted with ctx.post_message to the conversation that changed the skill, never into the model's context. A live SKILL.md is written only by `hermes court appeal apply` or Apply fix.Communitytools