Daily Todo Manager

Problem

• Engineers often scatter tasks across chats, notes, and project files, making it hard to reconstruct what happened after parallel sessions.
• Hand-maintained daily/projects files can overwrite timelines, omit completed work, or miss deadline and efficiency signals.
• Direct model writes can also clobber user-authored diary sections and historical records if file editing is too aggressive.
• This approach can reduce cross-session overwrites and missed project updates.

How It Works

• Organizes local data under TODO_DIR with daily/, projects/, and memory/, and uses project-tasks-cache.json to avoid scanning every project file.
• Phrases like “morning” generate today’s plan, “start task” predicts duration, and “end work” summarizes JSONL transcripts into timelines, completed tasks, reflections, and project logs.
• Writes emphasize Read then incremental Edit, preserving ## 🕐 Timeline, ## ✅ Completed, and user-written blocks; risk checks target deadlines within three days, overdue items, and efficiency anomalies.

Fit and Caveats

• Best for engineers who maintain long-running research or project tasks in local Markdown files.
• Requires TODO_DIR and CLAUDE_TRANSCRIPT_DIR; date interpretation, project ownership, and task classification still depend on triggers, keywords, and user confirmation.
• If the directory layout is missing or the cache is stale, initialize or rebuild the cache before writing logs.