lawve-ai/icc-jeanne-sulzer
Research, drafting, and analysis involving the International Criminal Court (Rome Statute system). Enforces a verification-first methodology — every case-law, decision, filing, warrant, and statement citation is verified against an authoritative primary source (icc-cpi.int, legal-tools.org) in the current conversation before it appears in an output. Foundational texts (Rome Statute, Elements of Crimes, RPE, Regulations of the Court) may be cited from project knowledge when present. Use whenever the user is working with ICC case law, OTP statements, ASP documents, or ICC procedure.
npx skills add https://github.com/lawve-ai/awesome-legal-skills --skill icc-jeanne-sulzer
For any case-specific document — judgment, decision, warrant, filing, OTP
statement — verify before citing. Verification means retrieving the
document from a Tier 1 source (icc-cpi.int, legal-tools.org, asp.icc-cpi.int)
in the current conversation. The four foundational texts (Rome Statute,
Elements of Crimes, Rules of Procedure and Evidence, Regulations of the
Court) are the only exception, and only when they are present in project
knowledge.
Nothing else may be cited from memory. Inventing a document number, a date,
or a paragraph reference is the single failure mode this skill exists to
prevent.
warrants, OTP statements, ASP resolutions, or ICC procedural texts.
note — that will reference ICC instruments or case law.
modes of liability, war crimes / crimes against humanity / genocide /
aggression elements, victim participation, reparations, or other
Statute-grounded questions.
If the question is about another tribunal (ICTY, ICTR, MICT, SCSL, STL,
ECCC, KSC, or hybrid mechanisms), this skill does not apply — use the
relevant tribunal's skill or, if none exists, surface that to the user.
The full procedure is in references/verification-workflow.md. The short
form:
user names a case or document, disambiguate against the document's own
header before proceeding. (Bemba TJ vs. Bemba AJ; Ntaganda 2017
interlocutory vs. 2021 final appeal — identity errors propagate.)
proposition each supports.
icc-cpi.int → legal-tools.org → search-result snippet from a Tier 1
domain → ICC press release → authoritative secondary database (e.g.
OUP ORIL) → ask the user. Stop at the first level that satisfies what
the claim needs.
(document, number, date, chamber), Content (the document holds, in
substance, what the output says it holds), Paragraph (the cited
paragraph contains the cited proposition). Pinpoint quotations require
paragraph-level verification. If verification stops short, soften the
claim or flag it explicitly.
references/citation-format.md.
or from this conversation's retrieval? Does the proposition match? Is
the verification level appropriate to the claim?
Read these as needed; they are the operational detail behind the workflow
above.
references/authoritative-sources.md — Tier 1 / Tier 2 / do-not-citesource hierarchy; the icc-cpi.int 403 fallback ladder.
references/citation-format.md — exact formats for the Rome Statute,Elements of Crimes, RPE, Regulations, decisions, warrants, OTP
statements, ASP documents. Includes the Article 28 numbering discipline
— the Rome Statute uses 28(a) (military commanders) and 28(b) (other
superiors), with no numbered paragraphs; the non-statutory 28(1)/(2)
form is the trap to avoid.
references/verification-workflow.md — the full operational procedure,the three-level verification gradient, a worked Bemba example.
references/foundational-texts.md — the four foundational texts andwhat is *not* foundational (case law, OTP policy papers, most ASP
resolutions, Regulations of the Registry/OTP, professional codes).
examples/example-verification.md — two end-to-end examples: fullparagraph-level verification (Bemba effective control) and partial
verification when icc-cpi.int blocks (Ntaganda 2017 jurisdiction).
examples/example-audit.md — two audit modes: working draft (the userwrote it; citations are claims to be checked) vs. finalised Court record
(the Court issued it; citations are part of the record, and the audit
is about how downstream work uses them).
who understand international criminal law.
workflow can produce accurate citations and disciplined drafts; only the
user can decide what to do with them.
These are non-negotiable. Edits or workarounds that weaken them defeat the
skill.
-Red) versions only in public outputs. Confidentialfilings (-Conf, -Conf-Exp) are not citable from a public output,
ever — even when the user has lawful access to the confidential version,
cite the -Red counterpart.
Article 28(a) (military commanders) and Article 28(b) (other
superiors) — not the non-statutory 28(1)/(2) form — in anything that
will be filed, read by judges or counsel, or quote the Statute.
are clearly separable in the output and never used to establish what
the Court has said or held.
do not abandon a real citation because the direct fetch was blocked.
Take lawve-ai/icc-jeanne-sulzer from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.