# Session Start > Start a work session - verify env, review progress, find next task Use when this capability is needed. - Skill: `tomevault-io/session-start-3` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/session-start-3` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/session-start-3/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Productivity - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/session-start-3 --- # Session Start Protocol Execute these steps in order: --- ## Step 1: Check Environment If a `package.json` exists, verify node_modules are installed: ```bash [ -f package.json ] && [ ! -d node_modules ] && npm install ``` If a `verify-env.sh` or similar script exists, run it. Report any environment issues before proceeding. --- ## Step 2: Read Recent Progress Check if `PROGRESS.md` exists at the project root. If it exists: - Read the last 50 lines to understand recent session context - Note the last session date and what was accomplished - Identify any "Next Session Should" guidance from the previous session If it doesn't exist: - Note: "No PROGRESS.md found - this may be a new project or first session" --- ## Step 2.5: Search Solutions Library Check for relevant prior learnings that might help with the current task. ### Determine Search Keywords From the previous session's notes or current task: - Extract domain keywords (e.g., "auth", "database", "api") - Extract technology keywords (e.g., "typescript", "prisma", "react") - Extract problem types (e.g., "build", "test", "performance") ### Search Project Solutions If `solutions/` directory exists: ```bash grep -r -l "[keyword]" solutions/ 2>/dev/null | head -5 ``` ### Search Global Solutions ```bash grep -r -l "[keyword]" ~/.claude/solutions/ 2>/dev/null | head -5 ``` ### If Matches Found Include in the session start summary: ``` **Relevant Prior Solutions:** - solutions/build-errors/typescript-path-aliases.md - ~/.claude/solutions/react/useeffect-cleanup.md ``` Brief summary of what each contains (read first few lines). ### If No Matches Skip this section silently — don't mention if nothing found. --- ## Step 3: Find Current Task Check if `ROADMAP.md` exists. If it exists: - Scan for unchecked items: `[ ]` (not `[x]`) - Find the first unchecked item as the likely "next task" - Note the section/phase it belongs to - Check for any items marked with `[WIP]` or similar in-progress indicators If it doesn't exist: - Note: "No ROADMAP.md found - consider creating one or using /roadmap skill" --- ## Step 4: Check Feature Status Check if `features.json` exists. If it exists: - Identify any features with status "wip" (work in progress) - Identify any features with status "blocked" - Note features that were recently completed (status "pass") If it doesn't exist: - Skip this step silently (features.json is optional) --- ## Step 5: Check for Blockers Look for: - Uncommitted changes: `git status --porcelain` - Failing tests (if test command exists in package.json) - TypeScript errors (if tsconfig.json exists): `npx tsc --noEmit 2>&1 | head -20` Report blockers that might affect the session. --- ## Output Format Produce a summary like this: ``` ## Session Start Summary **Last Session:** [date] - [brief summary from PROGRESS.md] **Previous Guidance:** [any "Next Session Should" notes] **Current State:** - Environment: [OK / issues found] - Uncommitted changes: [yes/no] - Tests: [passing / failing / not configured] **Next Task:** [description from ROADMAP.md] - Location: [section/phase in ROADMAP] - Context: [relevant background] **Ready to begin.** Say "let's start" or ask for clarification. ``` --- ## Graceful Handling If tracking files don't exist, don't error. Report what's missing and suggest: - "Run `/roadmap` to create a task breakdown" - "Create `PROGRESS.md` to track session history" This protocol should work for both established projects and new setups. --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/matthewod11-stack) — claim your Tome and manage your conversions.