# AI Resume Prompt — cg.cx Refinement Pass

## Mandatory First Steps on Every Resume
1. **Read `AI_MASTER_PLAN.md`** in full.
2. **Read `AI_CHECKPOINT.md`** in full.
3. **Read `AI_RESUME_PROMPT.md`** (this file) in full.
4. Continue from the exact batch listed in `AI_CHECKPOINT.md`.

## Execution Rules
- Follow the **exact batch order** in `AI_MASTER_PLAN.md`.
- Use the **mandatory 4-agent structure** for every batch:
  1. Security + Stability
  2. Telegram Bot + Permissions
  3. Content Delivery + Rendering
  4. Docs + QA + Regression
- Run the 4 subagents in **parallel**, wait for all to complete, then do a brief review.
- After review, **update `AI_CHECKPOINT.md`** with:
  - Completed tasks from the batch
  - Any blockers or deviations
  - Next batch to run
- **Do NOT skip batches.**
- **Do NOT merge batches.**
- **Do NOT do broad rewrites, redesigns, or cleanup-only changes.**

## Context Reminders
- This is a **refinement pass** on an existing, largely working Rust Telegram bot + web codebase.
- Many features (D, F, G, H, K, L, M, N, O, partial A/Q) are already implemented; verify they still work after each batch.
- The frontend is a Svelte 5 SPA with a retro theme. Preserve its style and behavior.
- SQLite is the database; do not add migrations unless absolutely necessary.
- The bot and server are separate binaries sharing the same DB and config.

## If Blocked
- If a subagent reports a blocker, pause the cycle.
- Update `AI_CHECKPOINT.md` with the blocker details.
- Do not proceed to the next batch until the blocker is resolved.