AI implementation roadmap
A phased plan for what to implement in this repo with strong AI-assist / agent fit: scripts, checks, skills, and generated docs. Use this to brief Cursor (or any agent) on the next buildable slice.
Principles
- Keep notes-first unless a phase explicitly adds runnable code.
- Prefer small PR-sized tasks an agent can finish in one session.
- Every phase lists exit criteria so “done” is unambiguous.
Phase 0 — Already shipped (baseline)
| Item | Status |
|---|---|
| Problem mind-maps + 12-section template + diagrams | Done |
| Patterns (20) + system design stack | Done |
scripts/validate-problems.sh | Done |
scripts/refresh-knowledgebase.sh + docs/AI-KNOWLEDGE-BASE.md + AGENTS.md snapshot | Done |
Skill: knowledgebase-refresh | Done |
Rule: dsa-mind-map-repo.mdc | Done |
Agent note: Do not redo Phase 0 unless migrating layout.
Phase 1 — Guardrails (highest ROI, low scope creep)
| ID | Implement | AI-fit | Exit criteria |
|---|---|---|---|
| 1.1 | GitHub Actions workflow | Agent writes YAML | On PR/push: ./scripts/validate-problems.sh exits 0; optional python3 -m compileall on scripts/*.py |
| 1.2 | Internal markdown link checker | Small Python in scripts/ | Reports broken ](*.md) relative links; CI fails on broken links |
| 1.3 | CONTRIBUTING.md | Agent drafts from templates | Single doc: clone, edit rules, diagram rule, run refresh-knowledgebase.sh, PR checklist |
| 1.4 | PROBLEM_INDEX vs disk diff | Script compares index rows to topics/**/problems/*.md | Prints missing/orphan files; exit 1 if orphan LC files not in index (policy configurable) |
Suggested order: 1.3 → 1.1 → 1.2 → 1.4.
Phase 2 — Knowledge automation
| ID | Implement | AI-fit | Exit criteria |
|---|---|---|---|
| 2.1 | Extend refresh script | Python | Optionally append “recent git diff summary” to docs/AI-KNOWLEDGE-BASE.md (last N commits touching topics/) |
| 2.2 | TAG_INDEX.md generator | Parse ## Mind-Map Tags | One page: tag → list of problem links |
| 2.3 | Spaced-repetition export | Script → CSV | Columns: front (title), back (recognition cues bullets), tags |
| 2.4 | Second Cursor skill: “new problem mind-map” | SKILL.md | Checklist: copy template, name file, fill 12 sections, run validate + refresh |
Suggested order: 2.4 → 2.2 → 2.3 → 2.1.
Phase 3 — Optional site and search
| ID | Implement | AI-fit | Exit criteria |
|---|---|---|---|
| 3.1 | MkDocs or VitePress | Agent scaffolds config | npm run docs:dev serves searchable sidebar by topic; deploy recipe in README |
| 3.2 | Algolia / Pagefind / built-in search | Config only | Search works on deployed site |
Dependency: Phase 1 link checker reduces broken links before publishing.
Phase 4 — Runnable solutions (explicit scope increase)
| ID | Implement | AI-fit | Exit criteria |
|---|---|---|---|
| 4.1 | solutions/go layout | One package, one LC smoke test | Example: two-sum with table-driven test from snippet |
| 4.2 | solutions/js layout | Same as Go | Jest or Node test runner documented |
| 4.3 | Do not mirror all 226 problems at once | Agent batches by topic | CONTRIBUTING explains “solutions optional per problem” |
Warning: Doubles maintenance whenever snippets change; use only if you want executable truth.
Phase 5 — Advanced quality
| ID | Implement | AI-fit | Exit criteria |
|---|---|---|---|
| 5.1 | Mermaid validation | CLI in CI or npx @mermaid-js/mermaid-cli | Fails if fenced mermaid blocks don’t parse |
| 5.2 | Normalize “At a Glance” | Scripted + manual review | Companies / Frequency use consistent enums |
| 5.3 | Duplicate snippet detector | AST-light overlap across Go blocks | Report only; human decides refactor |
Recommended AI sprint slices (1–2 hours each)
- Slice A:
CONTRIBUTING.md+ Phase 1.3 - Slice B:
.github/workflows/validate.yml+ Phase 1.1 - Slice C:
scripts/check_internal_links.py+ Phase 1.2 - Slice D:
scripts/diff_problem_index.py+ Phase 1.4 - Slice E: Skill
new-problem-mind-map+ Phase 2.4
Out of scope for this roadmap
- Replacing markdown with a database (unless you explicitly want an app).
- Auto-solving LeetCode via browser automation (fragile, ToS-sensitive).
- Full video course production.
How to use this doc with Cursor
Paste in chat: “Implement Phase 1.2 from the AI tooling roadmap” or “Do Slice B only.”
After each slice: run ./scripts/refresh-knowledgebase.sh and commit.
Related docs
- Further improvements backlog — broader ideas (human + tool mix)
- AI knowledge base snapshot — live inventory
AGENTS.md— handoff and conventions (repository root)
Last updated on
Spotted something unclear or wrong on this page?