소스 정보
- 저장소
- LangSensei/swat-marketplace
- 최근 소스 활동
- 2026년 4월 28일 15:23
- 감지된 SKILL.md 언어
- 영어
- 스타
- 1
- 포크
- 0
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
메뉴
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/LangSensei/swat-marketplace --skill debrief명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | debrief |
| version | 3.2.0 |
| description | Operation completion gate — notify the user or dispatch the next task |
| dependencies | {"skills":[],"mcps":["swat"]} |
Every operation must end with a debrief. Choose exactly one exit:
Never both. Never neither.
Use this flowchart to decide which exit to take. Squad-specific debrief rules in the MANIFEST always override this general guidance.
START
│
├─ Does your squad MANIFEST have "Debrief Rules (mandatory)" section?
│ YES → Follow those rules exactly. Stop here.
│ NO → Continue below.
│
├─ Did your operation produce work that needs follow-up by another squad?
│ YES → Exit 2 (Dispatch)
│ NO → Continue below.
│
├─ Did your operation complete its assigned task successfully?
│ YES → Exit 1 (Notify)
│ NO → Continue below.
│
├─ Did your operation fail in a way another squad could fix?
│ YES → Exit 2 (Dispatch) with failure context
│ NO → Exit 1 (Notify) with failure report
│
END
| Scenario | Exit | Target |
|---|---|---|
| PR opened, needs review | Dispatch | swat-review |
| PR reviewed, changes requested | Dispatch | swat-dev |
| PR reviewed, approved, zero comments | Notify | — |
| Analysis/research completed | Notify | — |
| Lint found errors, needs fix | Dispatch | swat-dev |
| All checks pass | Notify | — |
| Task failed, another squad can help | Dispatch | relevant squad |
| Task failed, no recovery path | Notify | — |
Send a concise notification to the user with your key findings.
Call the swat_notify MCP tool with your notification message:
swat_notify({"operation_id": "<your-operation-id>", "message": "your notification message"})
operation_id— read from OPERATION.md frontmatter. This enables the desktop notification to link directly toreport.htmlwhen clicked. Optional but recommended.
Good — leads with conclusion, includes data:
PR #42 on LangSensei/swat approved with no comments. Clean merge candidate.
Good — failure with context:
Build failed for feature/mcp-retry on LangSensei/swat. Go compilation error in mcp/client.go:128 — undefined reference to RetryConfig. This appears to require a type definition that wasn't included in the brief.
Bad — too vague:
Operation complete. Check the report for details.
When further work is needed, use the swat_dispatch MCP tool to hand off to the next squad.
Call the swat_dispatch MCP tool with a task brief:
swat_dispatch({"brief": "your dispatch brief", "details": "additional context, file paths, specifics"})
brief— concise one-line task description (used for classification). Keep it short and actionable.details— expanded context, constraints, file paths, code snippets, error messages. Put lengthy context here instead of cramming it intobrief.
Every dispatch brief must be self-contained — the receiving squad has no access to your operation context. Include:
Good — PR review dispatch (swat-dev → swat-review):
Review PR #47 on LangSensei/swat (branch: swat/20260415-abc12345).
Changes: Added retry logic to MCP client with exponential backoff. Modified mcp/client.go and mcp/client_test.go.
Files changed: mcp/client.go, mcp/client_test.go, mcp/config.go
Good — fix dispatch (swat-review → swat-dev):
Fix review comments on PR #47, LangSensei/swat. Branch: swat/20260415-abc12345. Resume the existing branch.
Fixes needed:
[blocking] mcp/client.go:142 — retry loop does not check context cancellation, risk of infinite retry on shutdown
[blocking] mcp/client.go:156 — error wrapping loses original error type, breaks errors.Is() checks downstream
[suggestion] mcp/config.go:28 — MaxRetries default of 10 seems high, consider 3 with longer backoff
PR context: Adds retry logic to MCP client for transient network failures.
Bad — missing context:
Fix the PR issues.