@laitszkin/apollo-toolkit 2.13.3 → 2.13.4

package/CHANGELOG.md

@@ -4,6 +4,14 @@ All notable changes to this repository are documented in this file.
 
 ## [Unreleased]
 
+## [v2.13.4] - 2026-04-05
+
+### Changed
+- Update `learn-skill-from-conversations` so it must inventory the current repository's existing skills first, weigh repeated user corrections and error-driven lessons more heavily, extract duplicated workflow fragments into shared skills when warranted, wrap repeatedly customized external skills, and keep project-specific tooling patterns in the owning project's `~/.codex/skills/`.
+
+### Fixed
+- Synchronize `package-lock.json` metadata with the current package version and CLI bin aliases before release publication.
+
 ## [v2.13.3] - 2026-04-05
 
 ### Removed

package/codex/learn-skill-from-conversations/README.md

@@ -10,7 +10,12 @@ This skill extracts the latest conversations from `~/.codex/sessions` and `~/.co
 - Stops immediately when there are no recent sessions
 - Cleans up `sessions` files older than 7 days after reading
 - Deletes `archived_sessions` files after reading them
+- Reads existing skills in the current working repository before proposing new ones
+- Prioritizes repeated user corrections, reported errors, tool failures, and reusable workflow gaps
+- Encourages extracting duplicated workflow fragments into shared skills when several skills need the same pattern
+- Wraps repeatedly customized external skills in a local skill when that produces a more reusable workflow
 - Defaults to creating a new skill unless strong overlap is confirmed
+- Keeps project-specific tool workflows out of the shared catalog and places them in the relevant project's `~/.codex/skills/`
 - Validates each changed skill with `quick_validate.py`
 
 ## Project Structure
@@ -40,7 +45,7 @@ python3 scripts/extract_recent_conversations.py --lookback-minutes 1440
 ```
 
 - If output is `NO_RECENT_CONVERSATIONS`, no action is required.
-- Otherwise, review extracted `[USER]` / `[ASSISTANT]` messages and apply updates through `skill-creator`.
+- Otherwise, review extracted `[USER]` / `[ASSISTANT]` messages, compare the lessons against existing skills in the current repository, and apply updates through `skill-creator`.
 
 ## License
 

package/codex/learn-skill-from-conversations/SKILL.md

@@ -15,9 +15,9 @@ description: Learn and evolve the local skill library from recent Codex conversa
 ## Standards
 
 - Evidence: Extract recent Codex session history first and derive reusable lessons only from actual conversation patterns.
-- Execution: Inventory the current skill catalog before editing, prioritize repeated requests, user corrections, reported errors, and post-completion follow-up asks that reveal missing closure, then prefer a focused update to the strongest related skill or create a new skill only when the overlap is weak.
-- Quality: Take no action when there are no recent sessions, avoid unrelated broad refactors, and validate every changed skill.
-- Output: Report the analyzed sessions, extracted lessons, created or updated skills, and the reasoning behind each decision.
+- Execution: Inventory the current working directory's existing skills before editing, prioritize repeated requests, user corrections, tool failures, logic bugs, architecture mismatches, documentation drift, and post-completion follow-up asks that reveal missing closure, then prefer a focused update to the strongest related skill or create a new skill only when the overlap is weak.
+- Quality: Take no action when there are no recent sessions, avoid unrelated broad refactors, keep shared skills cross-project reusable, route project-specific tooling patterns into the relevant project's `~/.codex/skills/`, and validate every changed skill.
+- Output: Report the analyzed sessions, extracted lessons, created or updated skills, shared-vs-project-specific placement decisions, and the reasoning behind each decision.
 
 ## Overview
 
@@ -46,10 +46,15 @@ python3 ~/.codex/skills/learn-skill-from-conversations/scripts/extract_recent_co
 ### 2) Derive reusable lessons
 
 - Identify repeated user needs, recurring friction, and repeated manual workflows.
+- Focus especially on repeated needs, repeated user corrections, and user-reported errors, then ask how a skill can prevent the same failure mode from recurring.
 - Give extra weight to moments where the user corrected the agent, rejected an earlier interpretation, or pointed out a missing preference or requirement.
 - Give extra weight to user-reported errors, regressions, or avoidable mistakes, then ask how a skill can prevent repeating that failure mode.
+- Treat tool-call failures, broken code paths, logic mistakes, weak architecture choices, and outputs that drifted from official documentation as valuable evidence when they expose a reusable missing guardrail or workflow.
 - Treat a user follow-up that asks for cleanup or an omitted finalization step immediately after the assistant reported completion as evidence that the workflow's done criteria were incomplete.
 - When that kind of follow-up recurs, tighten the owning skill's completion checklist before considering any new-skill extraction.
+- Even when a user request was highly specific, check whether the underlying workflow can be generalized into a reusable skill for the same class of tasks.
+- When multiple existing skills use a near-identical workflow fragment, consider extracting that fragment into a dedicated shared skill instead of leaving the duplication in place.
+- When an external skill is repeatedly used with the same user-specific customization layer, prefer wrapping it in a new local skill that encodes those standing conventions.
 - Ignore one-off issues that do not provide reusable value.
 - Distinguish between:
   - repeated trigger intent that deserves a new skill
@@ -58,7 +63,7 @@ python3 ~/.codex/skills/learn-skill-from-conversations/scripts/extract_recent_co
 
 ### 3) Decide new skill vs existing skill (default: new skill)
 
-- First read the relevant skills already present in the working repository so you do not create a duplicate under a different name.
+- First read the relevant skills already present in the current working directory repository (for example `apollo-toolkit`) so you do not create a duplicate under a different name.
 - Prefer creating a new skill.
 - Edit an existing skill only when the lesson is strongly related.
 - Treat relation as strong only when all three conditions hold:
@@ -67,12 +72,17 @@ python3 ~/.codex/skills/learn-skill-from-conversations/scripts/extract_recent_co
   - The update does not dilute the existing skill's scope.
 - When the recurring lesson is mainly about preventing a known mistake, prefer updating the existing skill that should have prevented it instead of creating a parallel skill.
 - When several skills repeat the same narrow workflow fragment, prefer extracting that fragment into a dedicated shared skill instead of copying the same guidance again.
+- When the strongest candidate is an external skill but the user repeatedly adds the same customization or policy layer, create a wrapper skill that calls that external skill while encoding the recurring local conventions.
+- Decide whether the lesson is cross-project reusable before choosing the destination:
+  - Cross-project reusable skills belong in the shared skill library.
+  - Project-specific workflows, especially ones tied to custom tools from a single repository, belong in that project's `~/.codex/skills/` directory instead of the shared catalog.
 - If uncertain, create a new skill instead of expanding an old one.
 
 ### 4) Apply changes through skill-creator
 
 - Explicitly follow `$skill-creator` workflow before editing skills.
-- For new skills, initialize with `~/.codex/skills/.system/skill-creator/scripts/init_skill.py`, then complete `SKILL.md` and required resources.
+- For new shared skills, initialize with `~/.codex/skills/.system/skill-creator/scripts/init_skill.py`, then complete `SKILL.md` and required resources.
+- For new project-specific skills, create or update them under the relevant project's `~/.codex/skills/` directory instead of the shared catalog.
 - For existing skills, make minimal focused edits and keep behavior consistent.
 
 ### 5) Validate every changed skill
@@ -87,10 +97,11 @@ python3 ~/.codex/skills/.system/skill-creator/scripts/quick_validate.py <skill-p
 
 ### 6) Report result
 
-- Summarize analyzed sessions, repeated needs, user corrections, error-driven lessons, created/updated skills, and why each decision was made.
+- Summarize analyzed sessions, repeated needs, user corrections, error-driven lessons, created/updated skills, placement decisions, and why each decision was made.
 
 ## Guardrails
 
 - Take no action when there are no sessions in the last 24 hours.
 - Avoid broad refactors across unrelated skills.
 - Avoid duplicate skills when an existing skill is strongly related.
+- Do not promote project-specific tool usage into the shared catalog unless the workflow is clearly reusable across repositories.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@laitszkin/apollo-toolkit",
-  "version": "2.13.3",
+  "version": "2.13.4",
   "description": "Apollo Toolkit npm installer for managed skill copying across Codex, OpenClaw, and Trae.",
   "license": "MIT",
   "author": "LaiTszKin",