메뉴
Shared deliverable-as-file output conventions for all BetterCallClaude commands. Defines the bcc-output folder structure, file naming, and chat summary template.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | output-conventions |
| description | Shared deliverable-as-file output conventions for all BetterCallClaude commands. Defines the bcc-output folder structure, file naming, and chat summary template. |
| tools | ["Read","Grep","Glob","Bash","WebSearch","WebFetch","mcp__bettercallclaude-http-swiss-caselaw__cite"] |
This document defines the standard output behavior for all BetterCallClaude commands that produce deliverables longer than approximately one page (~500 words).
Every command that produces a long output (memo, research, strategy, analysis, triage batch, translation, draft) MUST:
Short outputs (cite, validate, refine, version, help, privacy status, summarize with --short) may remain in chat.
<working folder>/bcc-output/
└── YYYY-MM-DD-<slug>/
├── 01-intake.md
├── 02-research-memo.md
├── 03-strategy.md
├── 04-adversarial-review.md
├── 05-draft-<document>.md (or .docx for redline)
└── sources.md
bcc-output is the default folder name. The user can override it in the bettercallclaude.local.md playbook under "Stile e formato" → output folder preference.YYYY-MM-DD-<short-description> format (e.g., 2026-06-10-nda-review-meier)./legal-5step phases. Commands run individually write only the relevant file(s) into the same structure.sources.md is always present when MCP servers were used: it lists every source consulted with verification date — this is the paper trail for due diligence..md (Markdown)..docx (tracked changes where applicable).When writing a file, display in chat:
**[Title of deliverable]** written to `bcc-output/YYYY-MM-DD-slug/filename.md`
[3-5 line summary of key findings / conclusions]
Full document: `bcc-output/YYYY-MM-DD-slug/filename.md`
Sources: `bcc-output/YYYY-MM-DD-slug/sources.md`
Commands that produce file output under this convention:
| Command | Output File(s) |
|---|---|
legal (multi-agent) | varies by workflow |
research | 02-research-memo.md |
strategy | 03-strategy.md |
draft | 05-draft-<doc>.md or .docx |
adversarial | 04-adversarial-review.md |
workflow | all pipeline files |
translate | translation-<doc>.md |
doc-analyze | analysis-<doc>.md |
nda-triage | nda-triage-<doc>.md (batch: summary + individual) |
precedent | precedent-chain-<topic>.md |
legal-5step | all 5 files + sources.md |
briefing | briefing-plan.md |
Commands that stay in chat: cite, validate, refine, summarize --short, version, help, privacy, start, doctor, federal (short), cantonal (short).
Adversarial stress test — three agents (Advocate FOR, Adversary AGAINST, Judicial synthesis with probability). Trigger when: user wants to stress-test a position, check weaknesses before filing, or requests adversarial review. Also option '4' in /legal menu. Do NOT trigger for: initial research (swiss-legal-research), strategy (swiss-legal-strategy), or drafting — this is a quality layer, not first-pass analysis.
Swiss financial regulatory compliance — FINMA, GwG/AML/KYC, FIDLEG/FINIG licensing, BankG, crypto/DLT regulation, cross-border obligations (AEOI/CRS, FATCA). Trigger when: user asks about FINMA licensing, AML/KYC compliance, financial services cross-border, crypto regulation, or fintech data processing. Do NOT trigger for: general corporate law, non-financial data protection (data-protection-law), tax (fiscal agent), or cantonal zoning.
Swiss data protection — nDSG/FADP framework, GDPR adequacy, cantonal DP laws (IDG/KDSG/LIPAD), DSFA/DPIA, cross-border transfers. Trigger when: user asks about nDSG/GDPR compliance, DPAs, data transfers, DPIA, data subject rights, or references FDPIC/nDSG/FADP. Do NOT trigger for: fintech regulatory compliance (use compliance-frameworks), document analysis (swiss-document-analysis), or privilege routing (privacy-routing).
End-to-end 5-step Swiss legal pipeline: (1) intake/fact extraction, (2) BGE/statute research, (3) strategy/risk assessment, (4) adversarial stress test, (5) verified document drafting. Trigger when: user asks for full analysis, 'run all steps', 'full pipeline', or 'end-to-end'. Do NOT trigger for: single-step tasks (use individual commands), citation-only work, or translation. Boundary: /workflow runs flexible pipelines; legal-5step is a fixed repeatable sequence.
Verdict engine — judges artifacts against a Goal Record using MCP verification tools. Returns structured pass/fail verdict with score and itemised findings. Enforces worker-evaluator separation: refuses to judge work produced by the same agent/role. Used by /legal-loop. Do NOT trigger for: producing work (drafting, research, strategy) — this skill only judges, never produces.
Swiss legal intake — transforms vague or complex queries into actionable execution plans. Two modes: **Refine** (single-domain, Socratic dialogue ≤3 rounds, structured prompt) and **Briefing** (multi-domain/multi-jurisdiction, specialist panel assembly, execution plan with checkpoints). Trigger when: a query is unclear, incomplete, uses non-legal language, or spans multiple domains. Supports --quick, --optimize, --skip-briefing. Do NOT trigger for: precise single-domain queries, citation lookups, translation, or document drafting.