@laitszkin/apollo-toolkit 2.13.1 → 2.13.2

package/AGENTS.md

@@ -16,7 +16,7 @@ This repository enables users to install and run a curated set of reusable agent
 - Users can investigate application logs, slice them to precise time windows, search by keyword or regex, and produce evidence-backed root-cause findings.
 - Users can answer repository-backed questions with additional web research when needed.
 - Users can commit and push local changes without performing version or release work.
-- Users can manage Codex user-preference memory by reviewing the last 24 hours of chats, storing categorized memory documents under `~/.codex/memory`, and syncing a memory index into `~/.codex/AGENTS.md`.
+- Users can manage Codex user-preference memory by reviewing the last 24 hours of chats, storing reusable preference-first memory documents under `~/.codex/memory`, and syncing a memory index into `~/.codex/AGENTS.md`.
 - Users can research a topic deeply and produce evidence-based deliverables.
 - Users can research the latest completed market week and produce a PDF watchlist of tradeable instruments for the coming week.
 - Users can turn a marked weekly finance PDF into a concise evidence-based financial event report.

package/CHANGELOG.md

@@ -4,6 +4,12 @@ All notable changes to this repository are documented in this file.
 
 ## [Unreleased]
 
+## [v2.13.2] - 2026-04-05
+
+### Changed
+- Update `codex-memory-manager` to require reusable, preference-first memory files built around a normalized `Scope / Preferences / Maintenance / Evidence notes` template instead of project- or incident-specific memory logs.
+- Add a bundled memory-file template reference plus focused template-structure tests so future updates keep the new memory format and de-projectification rules aligned.
+
 ## [v2.13.1] - 2026-04-05
 
 ### Fixed

package/codex/codex-memory-manager/README.md

@@ -1,11 +1,12 @@
 # codex-memory-manager
 
-Persist durable user preferences from recent Codex conversations into categorized memory files and a synchronized AGENTS index.
+Persist durable user preferences from recent Codex conversations into reusable, preference-first memory files and a synchronized AGENTS index.
 
 ## Highlights
 
 - Reads the last 24 hours of `~/.codex/sessions` and `~/.codex/archived_sessions`
 - Stores categorized preference memory under `~/.codex/memory/*.md`
+- Uses a normalized `Scope / Preferences / Maintenance / Evidence notes` memory template
 - Keeps a normalized memory index at the end of `~/.codex/AGENTS.md`
 - Adds new index entries automatically when new preference categories appear
 - Preserves the existing language already used in `~/.codex/AGENTS.md`
@@ -19,10 +20,14 @@ Persist durable user preferences from recent Codex conversations into categorize
 ├── LICENSE
 ├── agents/
 │   └── openai.yaml
+├── references/
+│   └── templates/
+│       └── memory-file.md
 ├── scripts/
 │   ├── extract_recent_conversations.py
 │   └── sync_memory_index.py
 └── tests/
+    ├── test_memory_template.py
     ├── test_extract_recent_conversations.py
     └── test_sync_memory_index.py
 ```
@@ -49,6 +54,12 @@ Refresh the AGENTS memory index after updating the memory files:
 python3 scripts/sync_memory_index.py --agents-file ~/.codex/AGENTS.md --memory-dir ~/.codex/memory
 ```
 
+Use the bundled memory template when creating or refactoring category files:
+
+```bash
+sed -n '1,200p' references/templates/memory-file.md
+```
+
 ## License
 
 MIT. See `LICENSE` for details.

package/codex/codex-memory-manager/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: codex-memory-manager
-description: Manage persistent Codex user-preference memory from recent conversation history. Use when users ask to learn from the last 24 hours of chats, update `~/.codex/AGENTS.md`, maintain `~/.codex/memory/*.md`, or sync new preference categories discovered in `~/.codex/sessions` and `~/.codex/archived_sessions`.
+description: Manage persistent Codex user-preference memory from recent conversation history. Use when users ask to learn from the last 24 hours of chats, update `~/.codex/AGENTS.md`, maintain reusable preference-first memory under `~/.codex/memory/*.md`, or sync new preference categories discovered in `~/.codex/sessions` and `~/.codex/archived_sessions`.
 ---
 
 # Codex Memory Manager
@@ -16,7 +16,7 @@ description: Manage persistent Codex user-preference memory from recent conversa
 
 - Evidence: Derive memory only from actual recent Codex conversations, and keep each stored preference tied to concrete chat evidence.
 - Execution: Extract the last 24 hours first, classify durable user preferences into memory files, then refresh the AGENTS index section.
-- Quality: Ignore one-off instructions, avoid duplicating categories, and preserve the existing language and tone already used in `~/.codex/AGENTS.md`.
+- Quality: Ignore one-off instructions, avoid duplicating categories, preserve the existing language and tone already used in `~/.codex/AGENTS.md`, and keep memory entries cross-project reusable, preference-heavy, and light on repository- or incident-specific detail.
 - Output: Report which sessions were reviewed, which memory categories were created or updated, and whether the AGENTS index changed.
 
 ## Goal
@@ -59,31 +59,38 @@ python3 ~/.codex/skills/codex-memory-manager/scripts/extract_recent_conversation
 
 - Store memory files under `~/.codex/memory/*.md`.
 - Reuse an existing category file when the new preference clearly belongs there.
-- Create a new category file when the recent chats introduce a distinct new class of preferences. Example: if the existing files are Rust-focused and recent chats introduce stable Java preferences, add a new Java-oriented category file and index it.
+- Create a new category file only when recent chats introduce a distinct reusable preference class that does not fit an existing file.
 - Keep filenames in kebab-case and scoped to a real category, for example:
-  - `architecture-preferences.md`
-  - `workflow-preferences.md`
-  - `java-preferences.md`
-- Use this normalized structure inside each memory file:
+  - `engineering-workflow.md`
+  - `assistant-style.md`
+  - `integration-and-deployment-preferences.md`
+- Keep categories organized by reusable preference type, not by repository, issue, feature name, or one-off incident.
+- When a file mixes too many unrelated preference types, split it by decision domain rather than by project.
+- Prefer wording that captures a reusable choice pattern such as `Prefer X when Y` or `Do not do Z when Q`.
+- Strip or generalize project-specific nouns, module names, branch names, issue numbers, and niche scenario labels unless they are required to explain the durable preference.
+- Keep evidence notes concise and factual; they should justify the preference without turning the memory file into a project log.
+- Use the normalized structure from `references/templates/memory-file.md` inside each memory file:
 
 ```md
-# Architecture Preferences
+# User Memory - Engineering Workflow
+
+Last curated: 2026-04-05 09:20 HKT
 
 ## Scope
-User preferences about system design, reuse, abstractions, and code organization.
+User preferences about how engineering tasks should be investigated, planned, implemented, verified, merged, and documented across repositories.
 
 ## Preferences
-- Prefer extending existing modules over parallel implementations.
-  - Applies when: adding adjacent behavior in an existing codebase.
-  - Evidence: repeated direction from recent Codex conversations reviewed on 2026-03-18.
-- Avoid speculative abstractions and over-engineering.
-  - Applies when: choosing between a focused edit and a broader refactor.
-  - Evidence: explicit repeated user guidance in recent sessions.
+- Prefer modular abstractions over duplicated logic.
+  - Applies when: designing or refactoring implementation structure.
+- Skip planning artifacts for clearly small, localized, low-risk changes.
+  - Applies when: assessing whether a task needs formal spec documents.
 
 ## Maintenance
-- Keep entries concrete and action-guiding.
-- Merge duplicates instead of restating the same preference.
-- Replace older statements when newer evidence clearly supersedes them.
+- Keep entries concrete, action-guiding, and reusable across repositories.
+- Move overlapping preferences to a better-matched memory file instead of keeping mixed categories.
+
+## Evidence notes
+- 2026-03-22 through 2026-04-04 repeated workflow corrections consistently reinforced scoped planning, approval gating, and architecture-aware implementation.
 ```
 
 ### 4) Refresh the AGENTS memory index at the end of `~/.codex/AGENTS.md`
@@ -123,4 +130,12 @@ python3 ~/.codex/skills/codex-memory-manager/scripts/sync_memory_index.py \
 - Do not store secrets, tokens, credentials, or personal data that should not persist.
 - Do not invent preferences when the evidence is weak or ambiguous.
 - Do not create duplicate categories when a current memory document already covers the same theme.
+- Do not create memory files organized around a specific repository, issue number, feature branch, or single operational incident unless the user explicitly wants that narrower scope.
+- Do not preserve project-specific wording when a more general preference statement would retain the useful lesson.
 - Do not rewrite unrelated parts of `~/.codex/AGENTS.md`; only manage the memory index block at the end.
+
+## References
+
+Load only when needed:
+
+- `references/templates/memory-file.md`

package/codex/codex-memory-manager/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Codex Memory Manager"
   short_description: "Persist user preferences from recent Codex chats"
-  default_prompt: "Use $codex-memory-manager to review the last 24 hours of Codex sessions, update ~/.codex/memory/*.md, and refresh the memory index at the end of ~/.codex/AGENTS.md."
+  default_prompt: "Use $codex-memory-manager to review the last 24 hours of Codex sessions, keep ~/.codex/memory/*.md reusable and preference-first with the normalized memory template, and refresh the memory index at the end of ~/.codex/AGENTS.md."

package/codex/codex-memory-manager/references/templates/memory-file.md

@@ -0,0 +1,21 @@
+# User Memory - [Category Name]
+
+Last curated: [YYYY-MM-DD HH:MM TZ]
+
+## Scope
+User preferences about [decision domain] across repositories and workflows.
+
+## Preferences
+- Prefer [reusable choice or behavior].
+  - Applies when: [general condition, not a single incident].
+- Avoid [reusable anti-pattern].
+  - Applies when: [general condition].
+
+## Maintenance
+- Keep entries preference-heavy and reusable across projects.
+- Remove or rewrite project names, issue numbers, branch names, and niche module details unless they are necessary to explain the preference.
+- Split the file when it starts mixing unrelated decision domains.
+- Replace stale or conflicting entries instead of keeping both.
+
+## Evidence notes
+- [YYYY-MM-DD through YYYY-MM-DD concise summary of the repeated conversation evidence that justified the preference.]

package/codex/codex-memory-manager/tests/test_memory_template.py

@@ -0,0 +1,37 @@
+#!/usr/bin/env python3
+"""Tests for the codex-memory-manager memory template reference."""
+
+from __future__ import annotations
+
+import unittest
+from pathlib import Path
+
+
+TEMPLATE_PATH = (
+    Path(__file__).resolve().parents[1]
+    / "references"
+    / "templates"
+    / "memory-file.md"
+)
+
+
+class MemoryTemplateTests(unittest.TestCase):
+    def test_template_contains_required_sections(self) -> None:
+        content = TEMPLATE_PATH.read_text(encoding="utf-8")
+
+        self.assertIn("# User Memory - [Category Name]", content)
+        self.assertIn("## Scope", content)
+        self.assertIn("## Preferences", content)
+        self.assertIn("## Maintenance", content)
+        self.assertIn("## Evidence notes", content)
+
+    def test_template_enforces_preference_first_generalization_rules(self) -> None:
+        content = TEMPLATE_PATH.read_text(encoding="utf-8")
+
+        self.assertIn("preference-heavy and reusable across projects", content)
+        self.assertIn("Remove or rewrite project names, issue numbers, branch names", content)
+        self.assertIn("Split the file when it starts mixing unrelated decision domains", content)
+
+
+if __name__ == "__main__":
+    unittest.main()

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@laitszkin/apollo-toolkit",
-  "version": "2.13.1",
+  "version": "2.13.2",
   "description": "Apollo Toolkit npm installer for managed skill copying across Codex, OpenClaw, and Trae.",
   "license": "MIT",
   "author": "LaiTszKin",