Install any skill in seconds. Free to start, no credit card required.
Get Started Free →lawve-ai/stl-lebanon-jeanne-sulzer
Verification-first methodology for the Special Tribunal for Lebanon. Citations are verified against the STL legacy archive and legal-tools.org. Covers the Ayyash et al. case, the two contempt cases (STL-14-05 New TV/Al-Khayat and STL-14-06 Akhbar Beirut/Al-Amin), and the 2011 terrorism decision. Research aid, not legal advice. Part of the open-source "Skills for International Justice" library — methodology: github.com/jeannesulzer/international-criminal-tribunals-skills
Passed 18 of 22 test cases with the skill, and 13 of 22 without it.
and 4 more in the full report
5 fixed13 passed either way4 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
Following the formal closure of the tribunal at the end of 2023, who manages the ongoing residual responsibilities, such as protecting witnesses and managing sentence enforcement?
| Checked | Without skill | With skill |
|---|---|---|
| States that a small residual mechanism handles post-completion functions under the supervision of the United Nations. | Missed | Passed |
The instructions your model reads when the skill loads.
This skill governs every output that touches the Special Tribunal for Lebanon. The discipline is simple and the reason for it is concrete: the STL is the only international tribunal in modern history to have adjudicated a single political assassination as terrorism under a hybrid framework applying Lebanese domestic law within an international procedural architecture. Its case numbers are exact, its in absentia trial framework is unusual, and its 2011 Interlocutory Decision on the Applicable Law is one of the most-cited (and most-contested) post-9/11 jurisprudential developments in international criminal law. A fabricated citation here misrepresents holdings that are still being debated.
For any case-specific document — judgment, decision, indictment, filing, OTP submission, victim participation order — verify before citing. "Verify" means web_fetch (or equivalent retrieval) to stl-tsl.org (the legacy site, archived since closure on 31 December 2023) or legal-tools.org in the current conversation. Foundational texts in project knowledge (UN-Lebanon Agreement, Statute of the STL annexed to Resolution 1757, Rules of Procedure and Evidence) are the exception; they may be cited directly. Nothing else.
In practice, retrieval to stl-tsl.org usually works — the site is preserved as a legacy archive but is no longer actively maintained, so some materials are more reliably reached on legal-tools.org. Three levels:
Label the level where relevant. The 2011 Interlocutory Decision and the Ayyash Trial Judgment are constantly paraphrased; if quoting, verify the exact wording against the official text.
Step 0 — Identify the document. Before anything else, distinguish:
The case identifier matters. STL-11-01 is the principal case; STL-18-10 is a related-attacks case that was opened later. Confusing them corrupts everything downstream.
Step 1 — Plan citations. List every citation that will appear in the output. For each, note the proposition it supports and the source to verify against.
Step 2 — Verify with the fallback ladder. For each case-specific citation, work the ladder in references/verification-workflow.md: stl-tsl.org → legal-tools.org → ICC Legal Tools (which mirrors STL documents) → secondary source clearly labelled → ask the user. Capture case number, document number, chamber, date, paragraph.
Step 3 — Draft using verified material. Use the citation format in references/citation-format.md. Where verification is partial, say so.
Step 4 — Self-audit. Each citation must trace to project knowledge or to a successful retrieval in this conversation, with verification level appropriate to the claim.
test cases passed, vs 13 / 22 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.