| name | dev-status-tracking |
| description | Teaches and implements context management for long-running development projects. Use when context loss becomes a problem or when setting up status tracking for AI-assisted workflows. |
| disable-model-invocation | true |
| model | sonnet |
| effort | medium |
| license | MIT |
Context Management
Why This Matters
When you work on something complex with AI, it forgets things. Every conversation has a token limit (~200,000 tokens), and when you hit it, earlier parts get compressed or lost. AI might forget components already built, decisions already made, and approaches that did not work.
The solution is simple: keep a status file for AI to read at the start of every development task. Sub-agents also help by running heavy work in separate chats with separate token limits, preserving the main conversation's budget.
Interaction Method
If AskUserQuestion is available, use it for all prompts below.
If not, present each question as a numbered list and wait for a reply before proceeding. For multiSelect questions, accept comma-separated numbers (e.g. 1, 3). Never skip or auto-answer without explicit user consent.
Step 0: Before starting
-
Announce your execution plan: Before doing anything, state what you will do in this activity: "Here's what I'm going to do: 1) set up a development status file for your project, 2) configure automatic update rules in CLAUDE.md, 3) implement the one-activity-per-chat pattern for context preservation, 4) share context survival strategies for long-running projects." This is a commitment device – harder to skip steps you just announced.
-
Conditional teaching: Ask the user if they are familiar with context management in AI-assisted development – why AI forgets things and how a status file helps. If yes, give a one-sentence refresher. If no, explain it in simple terms: every AI conversation has a memory limit, and when you hit it, earlier context gets compressed or lost – a status file is a persistent document AI reads at the start of every task so it knows what's been built, what's in progress, and what to avoid.
Required: ALWAYS ask the question, ALWAYS give the refresher when the user says yes. Never skip this step because the user "is a designer" or "already demonstrated familiarity earlier." Users want a memory refresh on every activity, including ones they know. Phrases like "I'll skip the explainer (you're a designer)" are forbidden — they signal the model has decided ON BEHALF OF the user that a refresher isn't needed. The user, not the model, decides what's redundant. The refresher takes one sentence; the cost is trivial; the value to a tired user mid-session is high.
-
Output presentation rule: Present output incrementally – one section at a time. After each section, discuss with the user, get their input, then move to the next. Never dump an entire deliverable at once.
-
Challenge ideas: After the user shares an idea or decision, challenge it – surface blind spots, edge cases, future implications. Then let the user decide with full perspective. This is not about being negative – it's about pressure-testing ideas so the user makes better decisions.
BLOCKING REQUIREMENT: Wait for the user to acknowledge the plan before proceeding to Step 1.
Step 1: Set Up the Status File
If the status file is being placed under .design-engineer-plugin/design/dev/ (e.g., .design-engineer-plugin/design/dev/status.md), ensure the parent directory exists first: run mkdir -p .design-engineer-plugin/design/dev (Bash). The plugin uses lazy folder scaffolding – folders are created by the skill that needs them, not upfront.
Create a development status file using the template from status-tracking-template.md. This file is separate from CLAUDE.md (separation of concerns) and should contain:
- Completed features – what is already built and working
- In-progress work – what is currently being developed
- Planned features – what comes next
- Critical warnings – approaches that failed, gotchas to remember
- Design system status – current token and component inventory
- Available assets – icons, images, and other resources
- Architecture decisions – key choices and their rationale
Place this file in a predictable location (e.g., dev-status/development-context.md or status.md at the project root).
Step 2: Configure Automatic Updates
Add a rule to the project's CLAUDE.md requiring AI to:
- Read the status file at the start of every development task
- Update the status file after completing major features or phases
- Record any new warnings or failed approaches
Note: AI can forget to track status even with rules in place. The user should nudge AI to update after every big development phase.
Step 3: Implement the One-Activity-Per-Chat Pattern
For browser-based AI tools (pre-development):
- Each activity gets its own chat to prevent context degradation
- Save deliverables between sessions (to project knowledge, files, or equivalent)
- Start each new chat by checking the status tracker
- In browser-based tools, saving deliverables to the project's knowledge must be done manually – AI cannot do this on its own
For Claude Code (development):
- Use sub-agents for heavy work (context analysis, implementation, auditing)
- Each sub-agent runs in a separate chat with its own token budget
- Results come back to the main conversation without burning its tokens
- Prefer manual compaction with specific instructions over auto-compaction when approaching limits
Step 4: Context Survival Strategies
Share the strategies from context-survival-guide.md:
When to Update Context
- After every feature AI implements
- After any critical bug is fixed
- Before working on anything unfamiliar
- When the project structure changes
- When new patterns or components are established
Handling Chat Compaction
- When approaching 200k tokens, AI automatically compresses earlier conversation
- This compression loses some details – prefer to avoid hitting the limit
- If compaction happens, ask AI to re-read the status file and CLAUDE.md
- For critical work, start a fresh conversation rather than continuing a compacted one
Recovering Lost Context
- The status file is the primary recovery mechanism
- GitHub commit history serves as a secondary record
- IDE Timeline feature can recover individual file changes
- Regular commits prevent catastrophic context loss
Content Integrity
- No fabrication: Only include content the user explicitly provided or that was read from an existing deliverable file. If you see a gap – a missing project detail, an unclear tracking need – ask via AskUserQuestion. Never fill gaps silently. Never invent project status, feature lists, or architecture decisions. Never attribute content to a deliverable you haven't Read.
- Read before reference: When referencing any previous deliverable in your output, you MUST Read the file first. Do not quote from memory – read the actual file and use its actual content.
Anti-slop Writing
Before generating any text for the deliverable, read anti-slop-writing.md and apply its rules. Scan your output before presenting it to the user.
Decision Hierarchy
- User's direct input – they decide what is important to track
- Project documentation – the status file reflects real project state
- AI suggestions – propose what to include in status updates
What Comes Next
After context management is set up, suggest dev-github-workflow for version control (another context preservation mechanism), or dev-agent-setup to set up sub-agents that help preserve the main conversation's token budget.
Resource Files
