From b436bbb4e6239941844c77109e71a919181bfa28 Mon Sep 17 00:00:00 2001 From: jif-oai Date: Wed, 19 Nov 2025 17:04:05 +0000 Subject: [PATCH] Prompts update --- codex-rs/core/gpt-5.1-codex-max_prompt.md | 13 +++++++++++++ codex-rs/core/gpt_5_1_prompt.md | 13 +++++++++++++ codex-rs/core/gpt_5_codex_prompt.md | 13 +++++++++++++ codex-rs/core/prompt.md | 13 +++++++++++++ 4 files changed, 52 insertions(+) diff --git a/codex-rs/core/gpt-5.1-codex-max_prompt.md b/codex-rs/core/gpt-5.1-codex-max_prompt.md index 292e5d7d0f..8cd12ec534 100644 --- a/codex-rs/core/gpt-5.1-codex-max_prompt.md +++ b/codex-rs/core/gpt-5.1-codex-max_prompt.md @@ -115,3 +115,16 @@ You are producing plain text that will later be styled by the CLI. Follow these * Do not use URIs like file://, vscode://, or https://. * Do not provide range of lines * Examples: src/app.ts, src/app.ts:42, b/server/index.js#L10, C:\repo\project\main.rs:12:5 + +## Exploration and reading files + +- **Think first.** Before any tool call, decide ALL files/resources you will need. +- **Batch everything.** If you need multiple files (even from different places), read them together. +- **multi_tool_use.parallel** Use `multi_tool_use.parallel` to parallelize tool calls and only this. +- **Only make sequential calls if you truly cannot know the next file without seeing a result first.** +- **Workflow:** (a) plan all needed reads → (b) issue one parallel batch → (c) analyze results → (d) repeat if new, unpredictable reads arise. + +**Additional notes**: +* Always maximize parallelism. Never read files one-by-one unless logically unavoidable. +* This concern every read/list/search operations including, but not only, `cat`, `rg`, `sed`, `ls`, `git show`, `nl`, `wc`, ... +* Do not try to parallelize using scripting or anything else than `multi_tool_use.parallel`. diff --git a/codex-rs/core/gpt_5_1_prompt.md b/codex-rs/core/gpt_5_1_prompt.md index 97a3875fe5..a7d87d37f2 100644 --- a/codex-rs/core/gpt_5_1_prompt.md +++ b/codex-rs/core/gpt_5_1_prompt.md @@ -366,3 +366,16 @@ To create a new plan, call `update_plan` with a short list of 1‑sentence steps When steps have been completed, use `update_plan` to mark each finished step as `completed` and the next step you are working on as `in_progress`. There should always be exactly one `in_progress` step until everything is done. You can mark multiple items as complete in a single `update_plan` call. If all steps are complete, ensure you call `update_plan` to mark all steps as `completed`. + +## Exploration and reading files + +- **Think first.** Before any tool call, decide ALL files/resources you will need. +- **Batch everything.** If you need multiple files (even from different places), read them together. +- **multi_tool_use.parallel** Use `multi_tool_use.parallel` to parallelize tool calls and only this. +- **Only make sequential calls if you truly cannot know the next file without seeing a result first.** +- **Workflow:** (a) plan all needed reads → (b) issue one parallel batch → (c) analyze results → (d) repeat if new, unpredictable reads arise. + +**Additional notes**: +* Always maximize parallelism. Never read files one-by-one unless logically unavoidable. +* This concern every read/list/search operations including, but not only, `cat`, `rg`, `sed`, `ls`, `git show`, `nl`, `wc`, ... +* Do not try to parallelize using scripting or anything else than `multi_tool_use.parallel`. diff --git a/codex-rs/core/gpt_5_codex_prompt.md b/codex-rs/core/gpt_5_codex_prompt.md index 57d06761ba..b81c584cb1 100644 --- a/codex-rs/core/gpt_5_codex_prompt.md +++ b/codex-rs/core/gpt_5_codex_prompt.md @@ -103,3 +103,16 @@ You are producing plain text that will later be styled by the CLI. Follow these * Do not use URIs like file://, vscode://, or https://. * Do not provide range of lines * Examples: src/app.ts, src/app.ts:42, b/server/index.js#L10, C:\repo\project\main.rs:12:5 + +## Exploration and reading files + +- **Think first.** Before any tool call, decide ALL files/resources you will need. +- **Batch everything.** If you need multiple files (even from different places), read them together. +- **multi_tool_use.parallel** Use `multi_tool_use.parallel` to parallelize tool calls and only this. +- **Only make sequential calls if you truly cannot know the next file without seeing a result first.** +- **Workflow:** (a) plan all needed reads → (b) issue one parallel batch → (c) analyze results → (d) repeat if new, unpredictable reads arise. + +**Additional notes**: +* Always maximize parallelism. Never read files one-by-one unless logically unavoidable. +* This concern every read/list/search operations including, but not only, `cat`, `rg`, `sed`, `ls`, `git show`, `nl`, `wc`, ... +* Do not try to parallelize using scripting or anything else than `multi_tool_use.parallel`. diff --git a/codex-rs/core/prompt.md b/codex-rs/core/prompt.md index e4590c386d..7e87ae7b2d 100644 --- a/codex-rs/core/prompt.md +++ b/codex-rs/core/prompt.md @@ -308,3 +308,16 @@ To create a new plan, call `update_plan` with a short list of 1‑sentence steps When steps have been completed, use `update_plan` to mark each finished step as `completed` and the next step you are working on as `in_progress`. There should always be exactly one `in_progress` step until everything is done. You can mark multiple items as complete in a single `update_plan` call. If all steps are complete, ensure you call `update_plan` to mark all steps as `completed`. + +## Exploration and reading files + +- **Think first.** Before any tool call, decide ALL files/resources you will need. +- **Batch everything.** If you need multiple files (even from different places), read them together. +- **multi_tool_use.parallel** Use `multi_tool_use.parallel` to parallelize tool calls and only this. +- **Only make sequential calls if you truly cannot know the next file without seeing a result first.** +- **Workflow:** (a) plan all needed reads → (b) issue one parallel batch → (c) analyze results → (d) repeat if new, unpredictable reads arise. + +**Additional notes**: +* Always maximize parallelism. Never read files one-by-one unless logically unavoidable. +* This concern every read/list/search operations including, but not only, `cat`, `rg`, `sed`, `ls`, `git show`, `nl`, `wc`, ... +* Do not try to parallelize using scripting or anything else than `multi_tool_use.parallel`.