Updated the tools section to clarify the available standard and special tools, including their usage and syntax. Removed deprecated tools and improved formatting for better readability.
Append two trailing spaces to each consecutive metadata line in the
header block of every codex file so the GitHub preview renders them
on separate lines instead of word-wrapping the whole block as one
paragraph. The last line before the blank-line/separator is left
alone since the blank line already breaks the paragraph.
- Surface the new per-model Codex CLI prompts (OpenAI/codex/) inline
with the matching GPT-5.4, 5.3, and 5.2 rows
- Replace dead OpenAI/codex-cli.md link with a row pointing at the
new codex/ folder plus the friendly/pragmatic personality files
- Repath OpenAI/gpt-5.3-codex.md to its new home in OpenAI/codex/
- Fix Misc/Notion AI.md link (file is now Notion-AI.md)
Replaces old OpenAI/codex-cli.md and OpenAI/gpt-5.3-codex.md with a
structured OpenAI/codex/ folder: one file per model (12 models) plus
four personality files extracted from model_messages.instructions_variables
(friendly + pragmatic, with gpt-5.2-codex's slightly divergent variants
kept separate).
- Bold GPT-5.4, GPT-5.3, Gemini 3.1 Pro, Claude Opus/Sonnet 4.6 as lead entries
- Move older models and variants (without-tools, GPT-4.x, Gemini 2.x) into collapsed details
- Add issue request link to Contributing section
- Update repo description to lead with GPT-5.4/5.3
- Replace minimal README with keyword-rich description and tables linking every prompt
- Remove abandoned claude.html and favicon.ico (moved to system prompt explorer project)
- Update repo description with model names for search visibility
- Swap low-value topics for high-traffic ones (grok, xai, perplexity, gpt-5, claude-code, system-prompt)
Updated reminders and guidelines for Claude's interactions, including handling of image requests, cyber warnings, and ethical considerations in conversations. Modified formatting for clarity.
Summary of all changes applied:
**Dates updated:** All instances now reflect March 30, 2026 (Monday) instead of March 11/12.
**Section 1.9 — Knowledge Cutoff:** Added the new paragraph about formulating search queries with the correct current year (the "latest iPhone 2025" vs "latest iPhone 2026" example).
**Section 2.6 — Memory Scope:** Changed from "within the current Project" to "outside of any Claude Project" and removed the Projects sub-bullet.
**Section 2.7 — Safety Reminders:** Expanded with the new language about instructions harmful to longterm wellbeing, identity stability, and character degradation over extended interactions.
**Section 5 — Past Chats:** Changed scope from "in a project" to "outside of any projects."
**Section 6.2 — User Preferences:** Expanded with fuller language including the "always"/"for all chats" qualifier, two additional examples ("I'm an architect" / "Fix this Python code"), and the paragraph about users not being able to see their own `<userPreferences>` content.
**Section 7.1 — Search Behaviors:** Added the UNRECOGNIZED ENTITY RULE paragraph.
**Section 7.3.1 — Copyright:** Added the expanded language about stringing together multiple small quotes and the "removing quotation marks does not make something a summary" clarification.
**Section 9 — Tools:** Removed `tool_search` (not present in current prompt). Added `recommend_claude_apps`. Added all 15 Figma MCP tools (Section 9.11). Fixed `gmail_create_draft` required fields to `["body"]` only. Fixed "Bob Iger" typo.
**Section 11.4.1 — API MCP Servers:** Added Canva and Figma to the connected servers list.
**Section 14.1 — Request Evaluation Checklist:** Added Step 0 ("Does this need a visual at all?"). Expanded Step 1 and Step 2 with additional routing examples.
**Section 16 — Available Skills:** Added file-reading and pdf-reading.
Ásgeir Thor Johnson