{"id":52115396,"url":"https://github.com/halfmoon-mind/fusion-council","last_synced_at":"2026-08-05T01:30:43.500Z","repository":{"id":366994508,"uuid":"1276697016","full_name":"halfmoon-mind/fusion-council","owner":"halfmoon-mind","description":"Deterministic multi-model planning \u0026 review councils for Claude Code (Claude role panel + GPT-5.5). Installable plugin + marketplace.","archived":false,"fork":false,"pushed_at":"2026-07-02T06:37:32.000Z","size":1861,"stargazers_count":4,"open_issues_count":1,"forks_count":1,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-07-02T07:26:16.401Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":"JavaScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/halfmoon-mind.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-22T07:59:26.000Z","updated_at":"2026-07-02T06:37:36.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/halfmoon-mind/fusion-council","commit_stats":null,"previous_names":["halfmoon-mind/fusion-council"],"tags_count":2,"template":false,"template_full_name":null,"purl":"pkg:github/halfmoon-mind/fusion-council","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/halfmoon-mind%2Ffusion-council","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/halfmoon-mind%2Ffusion-council/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/halfmoon-mind%2Ffusion-council/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/halfmoon-mind%2Ffusion-council/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/halfmoon-mind","download_url":"https://codeload.github.com/halfmoon-mind/fusion-council/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/halfmoon-mind%2Ffusion-council/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":36293557,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-07-20T02:08:10.276Z","status":"online","status_checked_at":"2026-08-04T02:00:06.901Z","response_time":57,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-08-05T01:30:42.749Z","updated_at":"2026-08-05T01:30:43.495Z","avatar_url":"https://github.com/halfmoon-mind.png","language":"JavaScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Fusion Council\n\n![Fusion Council hero illustration showing multiple read-only review perspectives converging into a synthesized report](assets/fusion-council-hero.png)\n\nTwo deterministic, read-only **councils** for Claude Code that vet your work with\nmore than one model family before you commit to it:\n\n- **`/fusion-council:fusion-plan \u003ctask\u003e`** — before a non-trivial change. A Claude\n  role panel (architect · skeptic · test-strategist · maintainer) plus **GPT-5.6 Sol**\n  deliberate in parallel; a judge compares them on merit and one plan is synthesized.\n- **`/fusion-council:fusion-review`** — after the change. One generalist Claude seat\n  plus **GPT-5.6 Sol** review your current `git diff`, judged and synthesized → one\n  findings report. This duo is the bench-validated default (matched recall on the\n  settled case set — n=1 hard case, so recall deltas are anecdote — at ~half the\n  clean-diff noise and −21% cost; `EVAL.md` §B5′); pass\n  `args: { roles: true }` to restore the role panel (skeptic · test-strategist ·\n  maintainer + GPT-5.6 Sol).\n\nBoth are **read-only**: panelists may read the repo but never edit it. Only you (the\nmain session) implement, and only after you approve.\n\n## Install\n\n```\n/plugin marketplace add halfmoon-mind/fusion-council   # or: /plugin marketplace add /path/to/this/repo\n/plugin install fusion-council@fusion-council\n```\n\nLocal testing without GitHub: point the marketplace at this directory directly\n(`/plugin marketplace add .` from the repo root).\n\n## Use\n\n```\n/fusion-council:fusion-plan add rate limiting to the upload endpoint\n# ...review the plan, approve, implement...\n/fusion-council:fusion-review\n```\n\n## The GPT-5.6 Sol seat (optional)\n\nThe non-Claude panel seat shells out to the local **Codex CLI** (`codex exec`,\nChatGPT subscription — no metered API). If `codex` isn't installed — or a run fails\nor times out — that seat drops out and the council runs Claude-only; every report's\n**Council Coverage** line says so, so reduced diversity is never silent. Install\nCodex CLI to enable it. The seat hands its prompt to `codex exec` on **stdin**\n(the `-` / piped form, so large tasks/diffs aren't capped by `ARG_MAX`), which\nneeds a Codex CLI recent enough to support it (verified on v0.142.0); an older\nCLI just shows up as UNAVAILABLE in Council Coverage.\n\n## Telemetry \u0026 self-evaluation\n\nEach run appends one compact stats row (the judge JSON, per-seat claim counts,\ncoverage — never full seat answers) to `~/.fusion-council/telemetry.jsonl`, so the\ncouncil's own value is measurable from real usage: per-seat surviving insights,\nFP rate, cross-family agreement, and whether you actually acted on the output\n(`eval/ack.sh y|n`). `eval/analyze.sh` prints the scoreboard. This is the ONE\nwrite the otherwise read-only workflows perform — pass `noTelemetry: true` to\ndisable it. The full validation methodology (verified literature, seeded-bug\nbenchmark design, ablation arms `noGpt`/`roles`, kill metrics, and the §B5′\nre-validation that made duo the review default) lives in `EVAL.md`.\n\n## Layout\n\n```\n.claude-plugin/plugin.json        plugin manifest\n.claude-plugin/marketplace.json   one-plugin marketplace (this repo)\nagents/                           5 read-only role agents (Read/Glob/Grep only)\nskills/fusion-plan, fusion-review entry points (call the workflows by path)\nworkflows/*.js                    the orchestration (bundled; called via ${CLAUDE_PLUGIN_ROOT})\nEVAL.md, eval/                    validation methodology + telemetry scoreboard scripts\n```\n\nWorkflows aren't an auto-loaded plugin component, so the skills invoke them by\n`scriptPath` under `${CLAUDE_PLUGIN_ROOT}` — they ship as bundled data, not as a\nregistered-by-name workflow.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhalfmoon-mind%2Ffusion-council","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fhalfmoon-mind%2Ffusion-council","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhalfmoon-mind%2Ffusion-council/lists"}