@laitszkin/apollo-toolkit 2.14.4 → 2.14.5

package/CHANGELOG.md

@@ -4,6 +4,11 @@ All notable changes to this repository are documented in this file.
 
 ## [Unreleased]
 
+## [v2.14.5] - 2026-04-08
+
+### Changed
+- Clarify `learn-skill-from-conversations`, `codex-memory-manager`, and `weekly-financial-event-report` so Codex automation runs must treat an explicit `Automation memory:` path in the prompt as authoritative and not rely on `$CODEX_HOME` shell expansion being available.
+
 ## [v2.14.4] - 2026-04-07
 
 ### Changed

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

@@ -15,7 +15,7 @@ description: Manage persistent Codex user-preference memory from recent conversa
 ## Standards
 
 - 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.
+- Execution: Extract the last 24 hours first, classify durable user preferences into memory files, then refresh the AGENTS index section; when a Codex automation prompt includes an explicit `Automation memory:` path, use that path directly for run-memory read/write instead of assuming `$CODEX_HOME` is set.
 - 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.
 
@@ -42,6 +42,7 @@ python3 ~/.codex/skills/codex-memory-manager/scripts/extract_recent_conversation
 - If output is exactly `NO_RECENT_CONVERSATIONS`, stop immediately and report that no memory update is needed.
 - Review every returned `[USER]` and `[ASSISTANT]` block before deciding that a preference is stable.
 - The extractor also cleans up stale session files after reading, matching the existing conversation-learning workflow.
+- If the prompt provides an explicit `Automation memory:` path, open that file first when present so the current curation run can extend the previous automation summary without relying on shell variable expansion.
 
 ### 2) Distill only stable user preferences
 
@@ -121,6 +122,7 @@ python3 ~/.codex/skills/codex-memory-manager/scripts/sync_memory_index.py \
   - which categories were created or updated
   - whether a new category was introduced
   - whether the AGENTS memory index changed
+- When an explicit `Automation memory:` path is provided in the task prompt, also write the concise run summary back to that exact file path and prefer it over `$CODEX_HOME/...` interpolation when the shell environment is incomplete.
 - When the user asks what memory exists or asks why a known preference was not mentioned, include the already-stored preferences that are directly relevant to the question instead of summarizing only newly added entries.
 - When a stable preference already existed and was still reinforced by recent chats, say that it remains stored and point to the category where it lives.
 - If no durable preferences were found, say so explicitly and avoid creating placeholder memory files.

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

@@ -15,7 +15,7 @@ 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 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.
+- 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; when a Codex automation prompt includes an explicit `Automation memory:` path, treat that path as the authoritative run-memory location instead of assuming `$CODEX_HOME` is available in the shell.
 - 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.
 
@@ -42,6 +42,7 @@ python3 ~/.codex/skills/learn-skill-from-conversations/scripts/extract_recent_co
 - Otherwise, review all `[USER]` and `[ASSISTANT]` blocks from each returned session.
 - The extractor reads both `~/.codex/sessions` and `~/.codex/archived_sessions`.
 - After extraction completes, it deletes `~/.codex/sessions` records older than 7 days and deletes all files under `~/.codex/archived_sessions`.
+- If the prompt provides an explicit `Automation memory:` path, read that file first when it exists so the current run can build on the prior automation summary even when shell variable expansion is unavailable.
 
 ### 2) Derive reusable lessons
 
@@ -98,6 +99,7 @@ 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, placement decisions, and why each decision was made.
+- When running inside a Codex automation thread, also write the concise run summary back to the explicit `Automation memory:` path from the prompt when one is provided; prefer that concrete path over `$CODEX_HOME/...` interpolation if the environment variable is unset.
 
 ## Guardrails
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@laitszkin/apollo-toolkit",
-  "version": "2.14.4",
+  "version": "2.14.5",
   "description": "Apollo Toolkit npm installer for managed skill copying across Codex, OpenClaw, and Trae.",
   "license": "MIT",
   "author": "LaiTszKin",

package/weekly-financial-event-report/SKILL.md

@@ -15,7 +15,7 @@ description: Read a user-specified PDF that marks the week's key financial event
 ## Standards
 
 - Evidence: Research only events explicitly marked in the source PDF plus clearly material breaking developments, and verify claims with current authoritative sources.
-- Execution: Read the PDF first, prefer `pdf` for extraction but fall back to the bundled macOS PDFKit extractor when local PDF tooling is missing, lock the research window, check for an existing report covering that same window before duplicating work, research each marked event, then hand the final briefing to `pdf` for rendering and QA with deterministic table-safe layout rules when needed.
+- Execution: Read the PDF first, prefer `pdf` for extraction but fall back to the bundled macOS PDFKit extractor when local PDF tooling is missing, lock the research window, check for an existing report covering that same window before duplicating work, research each marked event, then hand the final briefing to `pdf` for rendering and QA with deterministic table-safe layout rules when needed; when a Codex automation prompt includes an explicit `Automation memory:` path, reuse that concrete path for run-memory notes instead of assuming `$CODEX_HOME` resolves in the shell.
 - Quality: Keep the report concise, Chinese-compatible, explicit about source-versus-breaking events, conflicts, uncertainty, PDF font safety, and long-text table legibility.
 - Output: Save only the final standardized PDF under the month folder using the financial-event-report naming scheme.
 
@@ -43,6 +43,7 @@ Before writing, confirm these facts from the PDF, user context, or current sourc
 - Any user-specified geography, market, sector, or asset-class focus
 - Output directory if the user specified one
 - Language preference if different from the Chinese default
+- Automation memory path when the task prompt explicitly provides an `Automation memory:` line and the run needs to persist a concise automation summary
 
 Do not guess any input that materially changes the research window or report scope.
 
@@ -183,6 +184,12 @@ swift scripts/extract_pdf_text_pdfkit.swift /absolute/path/to/source.pdf
 - If the output fails visual QA, revise the content or PDF requirements and call the `pdf` skill again.
 - After the final PDF passes QA, require the `pdf` skill to delete all temporary screenshots before finishing.
 
+### 8) Persist automation run memory when applicable
+
+- If the task prompt includes an explicit `Automation memory:` path, read that file first when it exists so the run can build on the prior summary.
+- After the final report is generated, write a concise run note back to that exact path.
+- Prefer the explicit path from the prompt over `$CODEX_HOME/...` interpolation when the shell environment does not define `CODEX_HOME`.
+
 ## Standard Report Requirements
 
 The report must contain these sections in order: