boilerplate/.cursorrules at develop · podverse/boilerplate · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
# Boilerplate Repo Rules

## Stack

- Node.js 24+, TypeScript strict, npm workspaces

## Code Quality

- No `any` types
- **Strict equality**: Use `===` and `!==` only (no `==` or `!=`).
- **Avoid type assertions (`as`)**: Prefer proper types, optional chaining, type guards, or narrowing.

## Formatting

- **JavaScript/TypeScript/JSON/CSS**: Single quotes, semicolons, 2-space indent.
- Run `npm run lint:fix` or `npm run prettier:write` to auto-fix.

---

## LLM WORKFLOW

### Context Gathering

For substantial requests, ask: "What's the goal?" / "GitHub issue?"

### Scope Management

Warn if drifting: "This seems outside scope. Continue?"

### Agent/plan work: do not run tests

- **Do not run** test or verification commands (e.g. `make e2e_test_web`, `npm run test`) during agent or plan implementation.
- **Only instruct the user** to run those commands after your work is done; provide the exact command(s) in a fenced `bash` block. See **response-ending-make-verify** skill for which commands to suggest.
- **E2E verification commands:** Always give commands that **skip API tests** by default (do not add `E2E_API_GATE_MODE=on`). Only add `E2E_API_GATE_MODE=on` when the change affected API or integration tests. The Makefile default is `E2E_API_GATE_MODE=off`.

---

## HISTORY TRACKING (Essential)

**Applies to every file-modifying response in this repo** — no exception for path or task size. When in doubt, update history.

Before work: Check/create `.llm/history/active/[feature]/[feature]-part-01.md`

After changes, update with:

- Session date, user prompt (EXACT text), decisions, files

### Prompt Recording (CRITICAL)

**Always use the user's EXACT, FULL prompt text.** Never summarize or paraphrase.

- **`#### Prompt (Developer)`** — Manually typed by the user
- **`#### Prompt (Agent)`** — System-generated (e.g., clicking "Build" on a plan)

### Real-Time Capture

**If a response will modify files, log the prompt FIRST.**

- **Start of response**: Log prompt to `.llm/history/active/[feature]/[file].md`
- **End of response**: Update with files changed and key decisions

Skip logging for pure Q&A or explanations that don't change files.

End file-modifying responses with: **LLM History**: Updated `.llm/history/active/[feature]/[file].md`

---

## PLAN MANAGEMENT

**Plans must be under 300 lines.** Plans live in `.llm/plans/active/` (not `.cursor/plans/`).

If over 300: STOP, split into sub-plans, save in `.llm/plans/active/`, then work sequentially.