Skip to content

feat(request-basis): borrow Gutachten methods — generation engine, staged gates, interpretation hook#3

Open
lilialla wants to merge 1 commit into
mainfrom
feat/gutachten-method-borrow
Open

feat(request-basis): borrow Gutachten methods — generation engine, staged gates, interpretation hook#3
lilialla wants to merge 1 commit into
mainfrom
feat/gutachten-method-borrow

Conversation

@lilialla

Copy link
Copy Markdown
Owner

What

Adapt complementary methodology from Youchu-lawhub/gutachten-civil-case (Apache-2.0) into request-basis. All borrowed ideas are re-expressed in this skill's voice; every new file carries a source-acknowledgment line.

This skill is the "查字典" (curated dictionary) design; the borrowed skill is the "会造句" (generative) design. The biggest add fills our largest structural gap: a procedure to construct a review schema when no seed matches.

New references (6)

  • generation_protocol — build a four-gate review schema from the norm itself when neither a rule-obligation group nor a request-basis seed matches (查字典 → 会造句 fallback); all output labelled model_inference / pending_mcp.
  • staged_gates — fact–question–norm valve + two user-confirmation gates (candidate list, lifecycle outline) + per-stage violation self-check + parallel drafting with unified-premise injection. Trigger-scoped to structured analysis only.
  • interpretation_toolkit — indeterminate-concept (过错/合理/显失公平/善意) hook routing to interpretation ranks and handing off to the legal-interpretation-argument / analogical-reasoning skills.
  • proportionality — match argument depth to how contested each element is (◎○· markers + depth column); still visits every lifecycle gate.
  • output_precheck — pre-drafting gate for coverage, citation verification (司法解释 forced second check), delivery-format conformance, one-pass revision.
  • capability_slots — tool-to-slot mapping for portability; degradation routes through the source labels.

Edits

  • norm_decomposition — denial-vs-defense operational test (Q1/Q2 + burden axis + cross-stage mounting); burden-axis corrections for 1174 (免责事由, defendant's burden), 1173 (数额 reduction, not 责任成立), 182 (紧急避险 适当补偿).
  • classification — five-level norm retrieval + 18-class norm-basis taxonomy + 无因管理/不当得利 outer-order note.
  • SKILL.md — wire new references into Quick Protocol (steps 7, 13–16), Shared References, Global Rules.
  • resource_manifest.json — register six new references with fallbacks.

Verification

  • manifest references resolve 22/22; logical-id references no dangling.
  • skill-link-check --strict clean; deployed copy + Codex mirror byte-identical, no drift.
  • Reviewed by independent agents (legal-method, integration, adversarial): integration clean; four burden-axis corrections applied; no fabricated law.

License

Borrowed methodology only; source skill is Apache-2.0. Each new file acknowledges the source; no verbatim text copied.

🤖 Generated with Claude Code

…ged gates, interpretation hook)

Adapt complementary methodology from Youchu-lawhub/gutachten-civil-case (Apache-2.0),
re-expressed in this skill's voice with source acknowledgment in each file.

New references:
- generation_protocol: build a review schema from norms when no seed group/base
  matches (查字典 → 会造句 fallback)
- staged_gates: fact–question–norm valve, two user-confirmation gates, parallel
  drafting with unified-premise injection (trigger-scoped to structured analysis)
- interpretation_toolkit: indeterminate-concept hook routing to interpretation
  ranks and the legal-interpretation-argument / analogical-reasoning skills
- proportionality: match argument depth to how contested each element is
- output_precheck: pre-drafting gate for coverage, citation verification, format
- capability_slots: tool-to-slot mapping for portability

Edits:
- norm_decomposition: denial-vs-defense operational test (Q1/Q2 + burden axis +
  cross-stage mounting); 1174/1173/182 burden-axis corrections
- classification: five-level norm retrieval + 18-class norm-basis taxonomy
- SKILL.md: wire new refs into Quick Protocol, Shared References, Global Rules
- resource_manifest: register six new references with fallbacks

Verified: manifest references resolve 22/22; skill-link-check --strict clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants