woolfie1101/trinity-force

Plan any project as a stage x goal roadmap with hard per-goal acceptance gates, then run it hands-off — routing each goal to the best of three engines. Triggers — (a) PLAN/design ("break this into stages", "make a goal roadmap", "N stages x M goals", "design a step-by-step execution plan", "스테이지로 나눠줘", "goal 로드맵 만들어줘", "로드맵 설계해줘"); (b) RUN/execute ("run it hands-off", "keep going while I'm away", "process the goals autonomously", "orchestrate overnight", "continue the roadmap", "다음 goal 이어서", "무인으로 진행해", "알아서 계속 진행"); (c) RESEARCH ("research this", "gather sources on", "web search and summarize", "리서치 해줘", "자료 조사해줘") — routed to the agy engine. Uses a single-source-of-truth LEDGER.md to drain goals one at a time without polluting main context, routes engineering goals to the Codex MCP, creative/judgement goals to Claude Code, and research/web + multimodal + large-context + cross-check goals to agy (Google Antigravity CLI), verifies every goal's Definition of Done against first-party evidence before advancing, runs an independent review at every stage boundary, and halts only at genuine human-decision gates. On a machine that also has a general-purpose goal-orchestrator skill (e.g. goalpost), only ONE should be enabled (see "Migration"); a project-specific roadmap skill for the target repo takes precedence over this general one.

Claude Skills are markdown + scripts loaded into Claude Code; scanned with the same pipeline.

Source:

github.com/woolfie1101/trinity-force

Provenance:

Found by crawling (provenance, not a security rating)

We found this by crawling public sources and nobody has attested to it. That is how most of the registry gets here, and it is not a mark against the artifact — it only means no authenticated act of publication is on record. An artifact can be the vendor's own and still appear here.