Project intent map

What changed, why it changed, and what needs attention.

Hub is the human reading surface for Mainline intent memory. Use it to review pending work, understand file-level constraints, and orient around important decisions. Use the CLI when you need to act: start, seal, lint, sync, or retrieve context.

Current attention status Critical

Needs attention. 272 sealed intents, 0 open work, 4 proposed. Attention needed: 3 proposed waiting >72h; run proposal doctor, 12 uncovered commits.

Viewing synced Mainline state. Drafts: current worktree only; sibling drafts are local diagnostics until publish + sync.

Needs attention

Audit and improve Codex-facing repository guidance proposed for 47d — run mainline doctor --proposals · 47d · 2 files proposed 记录 stale proposed notes 修复缺口 proposed for 7d — run mainline doctor --proposals · 7d · 1 file proposed 添加 README terminal GIF demo 录制脚本 proposed for 5d — run mainline doctor --proposals · 5d · 2 files proposed 在中英文 README 主标语下加入演示视频 proposed, waiting <1h for review · 2 files proposed

Important decisions

Recent intents that recorded a real decision with rationale.

在中英文 README 主标语下加入演示视频 — README 中的视频嵌入方式: 使用 GitHub user-attachments 裸 URL，而不是把 mp4 文件提交进 ma… · z2z23n0
同步中英文 README 到 Git 原生意图记录定位 — README 主叙事: 以“AI 时代的 Git / Git for the AI era”作为开场定位，强调 Agent 决策和�… · z2z23n0
稳定 retrieval status reachability PBT — 如何修复 Full PBT 的 stale 分支不可达失败: 用四个显式 witness 锁定 retrieval status 的可… · z2z23n0
补齐 Pi hook PR 的 local-dev 与验证缺口 — Pi local-dev 的执行目录: 安装时把 Mainline repo root 写入生成的 `.pi/extensions/mainline.ts`，… · cat
添加 README terminal GIF demo 录制脚本 — demo aha moment: 用 Agent A/Agent B 的 invoice export 场景展示 proposed_overlap 在代码冲突前拦�… · z2z23n0

Decision hotspots

Files with the most intent history. These are areas with concentrated decisions, constraints, or repeated changes.

README.md 70 intents internal/engine/templates/agents-md.md 51 intents README.zh.md 45 intents internal/engine/engine.go 42 intents internal/cli/root.go 39 intents internal/engine/sync.go 38 intents skills/mainline/SKILL.md 38 intents AGENTS.md 30 intents

Long-term lifecycle signals

Good overall

Abandonment: 11.8% · Supersession: 0% · Review queue: 4 (oldest 47d)

Long-term trend, not current alert. Abandonment and supersession rates reflect rework / decision evolution — not per-person scores.