tink-harness 1.17.0 → 1.17.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.claude-plugin/plugin.json +1 -1
- package/CHANGELOG.md +4 -0
- package/README.ko.md +46 -0
- package/README.md +46 -0
- package/VERSIONING.md +1 -1
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -4,6 +4,10 @@ All notable changes to Tink are tracked here.
|
|
|
4
4
|
|
|
5
5
|
## Unreleased
|
|
6
6
|
|
|
7
|
+
## [1.17.1] - 2026-06-30
|
|
8
|
+
|
|
9
|
+
- README / README.ko.md에 Before→After 대비 블록과 run record·verify evidence 실물 예시 추가.
|
|
10
|
+
|
|
7
11
|
## [1.17.0] - 2026-06-30
|
|
8
12
|
|
|
9
13
|
- `loop-engineering` 하네스 추가: 측정 가능한 수용 신호(테스트·lint·build 통과, 커버리지, 벤치마크, 점수 등)를 향해 한 번에 한 병목씩 반복하고, 예산 초과 시 현재 상태·원인·다음 행동을 보고한다. 독립 평가(자기 채점 금지)와 반복 로그를 강제. `goal-checkpoint`(일반 다단계)·`bug-diagnosis-loop`(버그 진단)와 `use_when`에서 명시적으로 구분됨.
|
package/README.ko.md
CHANGED
|
@@ -83,6 +83,24 @@ $tink:cast 인증 모듈 리팩터링 # Codex
|
|
|
83
83
|
|
|
84
84
|
## 실제로 남는 것
|
|
85
85
|
|
|
86
|
+
**Tink 없이 (Claude Code / Codex 단독):**
|
|
87
|
+
|
|
88
|
+
```text
|
|
89
|
+
"인증 모듈 리팩터링"
|
|
90
|
+
→ 계획은 채팅 기록에만
|
|
91
|
+
→ 완료 기준 없음
|
|
92
|
+
→ 다음 세션에서 맥락 소실
|
|
93
|
+
```
|
|
94
|
+
|
|
95
|
+
**Tink 사용 후 (`/tink:cast 인증 모듈 리팩터링`):**
|
|
96
|
+
|
|
97
|
+
```text
|
|
98
|
+
→ .tink/current/contract.json — 완료 조건
|
|
99
|
+
→ plan.md / checks.md — 보이는 계획 + 검증 단계
|
|
100
|
+
→ /tink:verify — "된 것 같다"가 아닌 증거로 증명
|
|
101
|
+
→ .tink/runs/…md — 재사용 가능한 간결한 기록
|
|
102
|
+
```
|
|
103
|
+
|
|
86
104
|
사소하지 않은 작업마다 열어보고, diff하고, 커밋할 수 있는 평범한 파일이 남습니다:
|
|
87
105
|
|
|
88
106
|
```text
|
|
@@ -96,6 +114,34 @@ $tink:cast 인증 모듈 리팩터링 # Codex
|
|
|
96
114
|
refactor-review.md # 재사용 작업 방식 — 승인해야 저장
|
|
97
115
|
```
|
|
98
116
|
|
|
117
|
+
<details>
|
|
118
|
+
<summary><strong>완성된 run은 이렇게 남습니다</strong></summary>
|
|
119
|
+
|
|
120
|
+
**Run 기록** (`.tink/runs/YYYY-MM-DD-HHMM-작업명.md`):
|
|
121
|
+
|
|
122
|
+
```text
|
|
123
|
+
Status: completed
|
|
124
|
+
Goal: Codex에서 직접 호출 가능한 entrypoint 스킬 추가.
|
|
125
|
+
Changed: `$tink:<action>` 별칭을 인식하도록 main `tink` 스킬 수정; 래퍼 스킬 추가.
|
|
126
|
+
Evidence: `find … | rg 'tink(-|/)'` 로 모든 SKILL.md와 main 스킬 확인.
|
|
127
|
+
Notes: 재사용 메모리·하네스 변경 없음.
|
|
128
|
+
```
|
|
129
|
+
|
|
130
|
+
**Verify 증거** (`/tink:verify`, 두 가지 사례):
|
|
131
|
+
|
|
132
|
+
```text
|
|
133
|
+
✅ evidence_kind: command
|
|
134
|
+
evidence_ref: npm test
|
|
135
|
+
observed: 테스트 47개 통과, 실패 없음
|
|
136
|
+
|
|
137
|
+
⚠️ evidence_kind: manual
|
|
138
|
+
evidence_ref: 클린 설치 스모크 테스트 (macOS)
|
|
139
|
+
observed: 미실행 — CI 러너는 Linux 전용
|
|
140
|
+
next_action: 배포 전 macOS에서 수동 실행 필요
|
|
141
|
+
```
|
|
142
|
+
|
|
143
|
+
</details>
|
|
144
|
+
|
|
99
145
|
## CLAUDE.md·슬래시 명령·스킬만으로는 왜 부족할까?
|
|
100
146
|
|
|
101
147
|
| 도구 | 제공하는 것 | Tink가 얹는 것 |
|
package/README.md
CHANGED
|
@@ -83,6 +83,24 @@ $tink:cast refactor the auth module # Codex
|
|
|
83
83
|
|
|
84
84
|
## What you actually get
|
|
85
85
|
|
|
86
|
+
**Before (plain Claude Code / Codex):**
|
|
87
|
+
|
|
88
|
+
```text
|
|
89
|
+
"refactor the auth module"
|
|
90
|
+
→ plan lives in chat only
|
|
91
|
+
→ no completion criteria
|
|
92
|
+
→ context lost next session
|
|
93
|
+
```
|
|
94
|
+
|
|
95
|
+
**After (`/tink:cast refactor the auth module`):**
|
|
96
|
+
|
|
97
|
+
```text
|
|
98
|
+
→ .tink/current/contract.json — what must be true when done
|
|
99
|
+
→ plan.md / checks.md — visible plan and verification steps
|
|
100
|
+
→ /tink:verify — proves "done" with evidence, not vibes
|
|
101
|
+
→ .tink/runs/…md — compact record, reusable next time
|
|
102
|
+
```
|
|
103
|
+
|
|
86
104
|
Every non-trivial task leaves plain files you can open, diff, and commit:
|
|
87
105
|
|
|
88
106
|
```text
|
|
@@ -96,6 +114,34 @@ Every non-trivial task leaves plain files you can open, diff, and commit:
|
|
|
96
114
|
refactor-review.md # reusable ways of working — approval-gated
|
|
97
115
|
```
|
|
98
116
|
|
|
117
|
+
<details>
|
|
118
|
+
<summary><strong>What a finished run actually looks like</strong></summary>
|
|
119
|
+
|
|
120
|
+
**Run record** (`.tink/runs/YYYY-MM-DD-HHMM-task.md`):
|
|
121
|
+
|
|
122
|
+
```text
|
|
123
|
+
Status: completed
|
|
124
|
+
Goal: make Tink easier to invoke in Codex with direct entrypoint skills.
|
|
125
|
+
Changed: updated main `tink` skill to recognize `$tink:<action>` aliases; added thin wrapper skills.
|
|
126
|
+
Evidence: `find … | rg 'tink(-|/)'` showed all SKILL.md files plus the main skill.
|
|
127
|
+
Notes: no reusable memory or harness was changed.
|
|
128
|
+
```
|
|
129
|
+
|
|
130
|
+
**Verify evidence** (`/tink:verify`, two outcomes):
|
|
131
|
+
|
|
132
|
+
```text
|
|
133
|
+
✅ evidence_kind: command
|
|
134
|
+
evidence_ref: npm test
|
|
135
|
+
observed: 47 tests passed, 0 failures
|
|
136
|
+
|
|
137
|
+
⚠️ evidence_kind: manual
|
|
138
|
+
evidence_ref: clean install smoke (macOS)
|
|
139
|
+
observed: not run — CI runner is Linux only
|
|
140
|
+
next_action: run manually on macOS before publish
|
|
141
|
+
```
|
|
142
|
+
|
|
143
|
+
</details>
|
|
144
|
+
|
|
99
145
|
## Why not just CLAUDE.md / slash commands / skills?
|
|
100
146
|
|
|
101
147
|
| Tooling | What it gives you | What Tink adds on top |
|
package/VERSIONING.md
CHANGED