jhste-skills 0.14.2 → 0.14.4

package/CHANGELOG.md

@@ -1,5 +1,18 @@
 # Changelog
 
+## 0.14.4 - 2026-09-06
+
+### Changed
+- Updated `jhste-coding` to inspect the final task-owned diff for unnecessary, over-complex, or weakly-assumed parts, removing before simplifying, leaving sound work unchanged, and re-running any check the cleanup invalidates.
+- Synchronized both READMEs and released package version `0.14.4`.
+
+## 0.14.3 - 2026-09-05
+
+### Changed
+- Updated `jhste-coding` to report a pre-existing bug or unrelated problem noticed during the work as a follow-up instead of fixing it, unless the requested behavior cannot work without that fix.
+- Added an `Approaches tried and set aside` section to the portable handoff so the next executor does not repeat abandoned approaches.
+- Re-reviewed the Anthropic Claude Fable 5.1 guidance and general Claude prompting best practices, synchronized both READMEs, and released package version `0.14.3`.
+
 ## 0.14.2 - 2026-09-05
 
 ### Changed

package/README.en.md

@@ -16,14 +16,14 @@ Core `SKILL.md` files contain the task contract and important boundaries. Rare f
 ## Build and fix
 
 - **[`jhste-prototype`](skills/jhste-prototype/SKILL.md)** — tests one design question with the smallest runnable experiment, loads logic or UI detail only when relevant, and resolves an ambiguous question from the surrounding code with a recorded assumption when the user is unavailable.
-- **[`jhste-coding`](skills/jhste-coding/SKILL.md)** — implements clear features, known fixes, refactors, exact resume steps, and in-progress merge or rebase conflict resolution while preserving repository contracts, and prefers targeted edits over rewriting otherwise unchanged files when behavior is equivalent.
+- **[`jhste-coding`](skills/jhste-coding/SKILL.md)** — implements clear features, known fixes, refactors, exact resume steps, and in-progress merge or rebase conflict resolution while preserving repository contracts, prefers targeted edits over rewriting otherwise unchanged files when behavior is equivalent, reports unrelated problems noticed during the work as follow-ups instead of fixing them, and removes unnecessary or over-complex parts of its own change before reporting completion.
 - **[`jhste-diagnosing-bugs`](skills/jhste-diagnosing-bugs/SKILL.md)** — diagnoses uncertain failures and performance regressions through symptom-specific signals, competing explanations, runtime evidence, and measurement while redacting secrets from surfaced diagnostic evidence.
 
 ## Divide and continue work
 
 - **[`jhste-subagent-orchestration`](skills/jhste-subagent-orchestration/SKILL.md)** — chooses the lightest useful structure among direct work, one owner, parallel owners, and dynamic verified waves, and keeps independent head-owned work moving while workers are pending when the harness permits it.
 - **[`jhste-to-tickets`](skills/jhste-to-tickets/SKILL.md)** — turns defined work into one useful GitHub issue or an issue graph with real dependencies, drafting first and asking user-owned questions alongside the draft.
-- **[`jhste-handoff`](skills/jhste-handoff/SKILL.md)** — creates a portable transfer for another executor or a durable record for shared evolving state.
+- **[`jhste-handoff`](skills/jhste-handoff/SKILL.md)** — creates a portable transfer for another executor or a durable record for shared evolving state, including approaches tried and set aside so the next executor does not repeat them.
 
 ## Review and finish
 

package/README.md

@@ -16,14 +16,14 @@
 ## 직접 만들고 고치기
 
 - **[`jhste-prototype`](skills/jhste-prototype/SKILL.md)** — production 구현 전에 하나의 설계 질문을 가장 작은 실행 가능한 실험으로 검증하고, 필요한 logic 또는 UI 지침만 추가로 읽으며, 질문이 모호하고 사용자가 없으면 주변 코드에 맞는 해석을 택하고 그 가정을 기록합니다.
-- **[`jhste-coding`](skills/jhste-coding/SKILL.md)** — 충분히 이해된 기능, 알려진 수정, 리팩터링, 정확한 재개 지점, 진행 중인 merge/rebase conflict를 저장소 계약에 맞게 처리하며, 동작이 같다면 변경되지 않은 파일 전체를 다시 쓰기보다 필요한 부분만 편집합니다.
+- **[`jhste-coding`](skills/jhste-coding/SKILL.md)** — 충분히 이해된 기능, 알려진 수정, 리팩터링, 정확한 재개 지점, 진행 중인 merge/rebase conflict를 저장소 계약에 맞게 처리하며, 동작이 같다면 변경되지 않은 파일 전체를 다시 쓰기보다 필요한 부분만 편집하고, 작업 중 발견한 무관한 문제는 고치지 않고 후속 과제로 보고하며, 완료 보고 전 자신이 만든 변경에서 불필요하거나 과하게 복잡한 부분을 걷어냅니다.
 - **[`jhste-diagnosing-bugs`](skills/jhste-diagnosing-bugs/SKILL.md)** — 원인이 불명확한 장애와 성능 저하를 증상별 신호, 경쟁 설명, runtime 증거, 측정으로 진단하며 노출되는 진단 자료의 secret을 가립니다.
 
 ## 일을 나누고 이어가기
 
 - **[`jhste-subagent-orchestration`](skills/jhste-subagent-orchestration/SKILL.md)** — 직접 수행, 단일 소유, 병렬 소유, 동적 wave 가운데 작업에 맞는 가장 가벼운 구조로 워커를 조정하고, 하네스가 허용하면 워커 결과와 독립적인 head 작업은 기다리지 않고 계속합니다.
 - **[`jhste-to-tickets`](skills/jhste-to-tickets/SKILL.md)** — 정의된 작업을 하나의 유용한 GitHub issue 또는 실제 dependency가 있는 issue graph로 나누고, 사용자 결정이 필요한 부분은 초안을 먼저 만든 뒤 초안과 함께 묻습니다.
-- **[`jhste-handoff`](skills/jhste-handoff/SKILL.md)** — 다른 실행자로 넘기는 portable handoff 또는 여러 소유자가 유지하는 durable handoff를 만듭니다.
+- **[`jhste-handoff`](skills/jhste-handoff/SKILL.md)** — 다른 실행자로 넘기는 portable handoff 또는 여러 소유자가 유지하는 durable handoff를 만들며, 시도했다가 버린 접근과 그 이유를 남겨 같은 막다른 길을 반복하지 않게 합니다.
 
 ## 검토하고 마무리하기
 

package/THIRD_PARTY_NOTICES.md

@@ -104,11 +104,17 @@ The 0.14.1 maintenance pass reviewed Anthropic's public prompting guidance for C
 - when the harness allows the head to keep working while delegated workers are pending, continue safe head-owned work that does not depend on those results and wait only at a real dependency barrier;
 - when behavior is equivalent, prefer a targeted edit over rewriting otherwise unchanged files.
 
-Guidance reviewed on 2026-09-03:
+The 0.14.3 maintenance pass re-reviewed the Claude Fable 5.1 guidance together with the general Claude prompting best practices and adapted two more behaviors:
 
-- https://platform.claude.com/docs/ko/build-with-claude/prompt-engineering/prompting-claude-fable-5-1
+- `jhste-coding` reports a pre-existing bug or unrelated problem noticed during the work as a follow-up instead of fixing it, unless the requested behavior cannot work without that fix;
+- the portable handoff records approaches tried and set aside, mirroring what a compaction summary must preserve so the next executor does not repeat them.
 
-No guide text is copied. Model-specific effort, search-trigger, conversation-history, thinking-display, and API-message guidance remains outside these harness-neutral skills.
+Guidance reviewed on 2026-09-03 and 2026-09-05:
+
+- https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5-1
+- https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices
+
+No guide text is copied. Model-specific effort, search-trigger, conversation-history, thinking-display, progress-update, formatting, and API-message guidance remains outside these harness-neutral skills.
 
 ## OpenAI GPT-6 Astra prompting guidance
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jhste-skills",
-  "version": "0.14.2",
+  "version": "0.14.4",
   "description": "A lean, harness-neutral engineering skill set for autonomous coding, diagnosis, prototyping, subagent coordination, specifications, tickets, handoffs, implementation finalization, pull request review, decision interviews, external questionnaires, and domain modeling.",
   "type": "module",
   "license": "MIT",

package/skills/jhste-coding/SKILL.md

@@ -23,6 +23,8 @@ When behavior is equivalent, prefer targeted edits over rewriting otherwise unch
 
 Add an abstraction only when the current change demonstrates real variation, repeated change, or a clearer side-effect boundary. Avoid speculative extension points, pass-through wrappers, broad cleanup, and prototype scaffolding in production code.
 
+Report a pre-existing bug, performance concern, or unrelated problem noticed during the work as a follow-up in the completion report instead of fixing it, unless the requested behavior cannot work without that fix.
+
 For an in-progress merge or rebase conflict, read [references/merge-conflicts.md](references/merge-conflicts.md) before resolving the conflicted state.
 
 Continue through in-scope local edits without pausing. Stop only when progress requires a consequential user-owned decision, broader authority, an external or destructive action not already authorized, unresolved ownership conflict, or root-cause work that should move to diagnosis.
@@ -31,7 +33,7 @@ Continue through in-scope local edits without pausing. Stop only when progress r
 
 Use the strongest available repository-native signal that directly distinguishes the requested behavior from failure. Expand validation only when the change's risk, integration surface, or an observed failure justifies it; do not run a fixed test, type, lint, build, and smoke sequence by habit.
 
-Add or update a test only at a seam that represents the real caller-visible behavior. Inspect the final task-owned diff for scope creep, temporary instrumentation, stale compatibility paths, and prototype-only artifacts. Never imply that an unrun check passed.
+Add or update a test only at a seam that represents the real caller-visible behavior. Inspect the final task-owned diff for scope creep, temporary instrumentation, stale compatibility paths, prototype-only artifacts, and anything unnecessary, over-complex, or resting on an assumption the evidence does not support; remove before simplifying, leave what already reads well unchanged, and re-run any check the cleanup invalidates. Never imply that an unrun check passed.
 
 ## Completion
 

package/skills/jhste-handoff/references/portable.md

@@ -13,6 +13,8 @@ Include only material sections:
 
 ## Settled decisions
 
+## Approaches tried and set aside
+
 ## Authoritative references
 
 ## Changed or owned resources