investigatelisted

# Investigation — From Idea to Plan ## Goal: "$ARGUMENTS" Start a structured investigation for new work. The user provides a goal and hints about where to find relevant information. Claude gathers context, synthesizes understanding, creates a formal plan, and optionally files GitHub issues. This is the **beginning-of-work** counterpart to `/fullpostmortem` (end-of-work). ## Phase 1: Gather **Goal**: Build a complete picture of the problem space from all available sources. ### 1a. Parse User Input Extract from the user's message: - **Goal statement**: What are we trying to accomplish? - **File hints**: Any `*.md` filenames, paths, or document references → read them - **Issue hints**: Any `#N` references or issue descriptions → fetch with `gh_issue_full.py --full` - **Keywords**: Technical terms, feature names, component names → search for related docs - **Scope signals**: "small fix", "new feature", "refactor", "epic" → calibrate depth ### 1b. Read Referenced Sources (parallel where possible) Launch up to 3 agents in parallel based on what was referenced: **Local files** (most common starting point): - Read all files the user mentioned by name - Search `private/claude/` for related design docs (grep by keywords from goal) - Check for existing plans in `~/.claude/plans/` **GitHub issues** (CRITICAL — always chase references): - For each referenced issue: `python scripts/gh_issue_full.py N --full --repo OWNER/REPO` - Look for `gh_issue_full.py` in project's `scripts