Skip to content

Commit e90ca10

Browse files
author
DavidQ
committed
Level 11.1: tighten numeric validation at transition boundary
- enforce finite numeric validation using Number() normalization - prevent NaN / Infinity drift - improve validation clarity Scope: validation only
1 parent 68454bd commit e90ca10

7 files changed

Lines changed: 127 additions & 0 deletions

‎docs/dev/DAILY_STARTUP_ONE_SHOT.md‎

Lines changed: 22 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,22 @@
1+
# Daily Startup (One-Shot)
2+
3+
## ChatGPT Project
4+
Start a NEW chat in your dedicated Project and paste:
5+
6+
Use docs/dev/PROJECT_INSTRUCTIONS.md as system instructions.
7+
8+
Run full workflow for <PLAN_PR_NAME>
9+
10+
## What ChatGPT should return
11+
- PLAN validation result
12+
- compact BUILD_PR
13+
- Codex command
14+
- downloadable repo-structured ZIP
15+
16+
## Codex
17+
Run the generated command. Codex will rely on:
18+
- AGENTS.md
19+
- .codex/skills/repo-build/SKILL.md
20+
21+
## Why this works
22+
Projects keep the workflow context together, which is useful for repeated long-running work. AGENTS.md is auto-read by Codex before it starts, and skills load on demand rather than every session.

‎docs/dev/ONE_SHOT_MODE.md‎

Lines changed: 29 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,29 @@
1+
# One-Shot Mode
2+
3+
## Goal
4+
Let ChatGPT run the full non-code workflow from a single prompt:
5+
- validate PLAN
6+
- create BUILD
7+
- generate Codex command
8+
- package ZIP
9+
10+
## Trigger Prompt
11+
Run full workflow for <PLAN_PR_NAME>
12+
13+
Example:
14+
Run full workflow for PLAN_PR_LEVEL_11_2_RECONCILIATION_LAYER_FOUNDATION
15+
16+
## Expected ChatGPT behavior
17+
1. Read the referenced PLAN_PR
18+
2. Validate that the PLAN is singular, specific, and executable
19+
3. Create a compact BUILD_PR for the same scope
20+
4. Create/update the Codex command
21+
5. Package a repo-structured ZIP under <project folder>/tmp/
22+
6. Return the ZIP plus a short summary
23+
24+
## Boundaries
25+
- ChatGPT does docs and ZIP packaging only
26+
- Codex writes implementation code
27+
- One PR purpose only
28+
- No vague language
29+
- No repo-wide scanning unless exact targets require it

‎docs/dev/PROJECT_INSTRUCTIONS.md‎

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -26,3 +26,20 @@ Do not:
2626
- Write implementation code unless explicitly asked
2727
- Expand scope beyond the PR
2828
- Modify start_of_day folders unless requested
29+
30+
NEXT resolution rules:
31+
32+
If the user says "NEXT":
33+
1. Look for the highest completed or referenced PLAN_PR in the session
34+
2. Increment to the next logical PLAN_PR in sequence
35+
3. If sequence is unclear, STOP and ask for clarification
36+
37+
Assume naming pattern:
38+
PLAN_PR_LEVEL_<major>_<minor>_<name>
39+
40+
Example:
41+
If last = PLAN_PR_LEVEL_11_1_...
42+
NEXT = PLAN_PR_LEVEL_11_2_...
43+
44+
If no prior context exists:
45+
STOP and ask: "What is the base PLAN_PR?"

‎docs/dev/PROJECT_V2_NOTES.md‎

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,17 @@
1+
# Project V2 Notes
2+
3+
## Recommended operating model
4+
- Use a dedicated ChatGPT Project for this repo workflow
5+
- Start a NEW chat for each PR to keep the browser responsive
6+
- Use one-shot mode to reduce daily startup steps
7+
- Keep GitHub connector usage narrow and explicit
8+
9+
## Good GitHub prompts
10+
- Show me files in repo <repo-name>
11+
- Open file <path> in repo <repo-name>
12+
- Review PR <number> in repo <repo-name>
13+
14+
## Do not use GitHub for
15+
- broad repo scans
16+
- replacing local Codex execution
17+
- daily startup context

‎docs/dev/templates/BUILD_FROM_PLAN_TEMPLATE.md‎

Lines changed: 33 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,33 @@
1+
# BUILD_PR_<NAME>
2+
3+
## Purpose
4+
<single-sentence purpose copied and tightened from PLAN>
5+
6+
## Scope
7+
Primary target files:
8+
- <exact target 1>
9+
- <exact target 2>
10+
11+
Allowed nearby reads:
12+
- <small related area only if required>
13+
14+
## Required implementation
15+
- <exact change 1>
16+
- <exact change 2>
17+
18+
## Acceptance criteria
19+
- <criterion 1>
20+
- <criterion 2>
21+
22+
## Validation
23+
Run only:
24+
- <command 1>
25+
- <command 2>
26+
27+
## Non-goals
28+
- no unrelated refactor
29+
- no repo-wide scan
30+
- no tooling changes unless explicitly required
31+
32+
## Working tree rule
33+
If the tree is already dirty, ignore unrelated files and modify only the scoped files for this PR purpose.

‎docs/dev/templates/CHATGPT_ONE_SHOT_PROMPT.md‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,3 @@
1+
Use docs/dev/PROJECT_INSTRUCTIONS.md as system instructions.
2+
3+
Run full workflow for <PLAN_PR_NAME>

‎docs/dev/templates/CODEX_ONE_SHOT_COMMAND.md‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,6 @@
1+
MODEL: GPT-5.3-codex | REASONING: high
2+
3+
Use skill: repo-build
4+
TASK: <BUILD_PR_NAME>
5+
INPUTS: docs/pr/<PLAN_PR_NAME>.md, docs/pr/<BUILD_PR_NAME>.md
6+
OUTPUT: <project folder>/tmp/<BUILD_PR_NAME>_delta.zip

0 commit comments

Comments
 (0)