# Verification workflow — KSC The step-by-step procedure for verifying a KSC citation before it appears in an output. ## The principle No case-specific citation appears in an output until it has been verified against an authoritative source in the current conversation. Foundational instruments (Law 05/L-053, Article 162 Constitution, RPE, Host State Agreement) may be cited from project knowledge when present. ## The fallback ladder 1. **scp-ks.org** — the official KSC website. Actively maintained. First and primary source. 2. **legal-tools.org** — ICC Legal Tools Database. Mirrors KSC principal decisions. 3. **Council of Europe** (assembly.coe.int) — for the Marty Report and PACE Resolutions. 4. **Tier 2 scholarship** — for context and doctrinal commentary; never for tribunal findings. 5. **Ask the user** if nothing verifies. ## Capture these fields - **Case number** — KSC-BC-2020-04, KSC-BC-2020-05, KSC-BC-2020-06, etc. - **Document type** — Indictment (confirmed/amended), Decision, Trial Judgment, Sentencing Judgment, Reparation Order, Appeal Judgment - **Panel/chamber** — Pre-Trial Judge, Trial Panel I, Trial Panel II, Court of Appeals Panel, Supreme Court Panel, Constitutional Court Panel - **Date** — exact date of issuance - **Paragraph number(s)** — for pinpoint citations - **Public vs confidential / redacted version** — use public versions; the operative indictment in Thaçi et al. is the **lesser-redacted version of 27 February 2023** ## Match the verification level to the claim - "Mustafa was convicted by the KSC Trial Panel on 16 December 2022" → existence verified - "The Trial Panel held that the killings at Zllash were part of a widespread or systematic attack…" → content verified - A direct quotation from the Mustafa Trial Judgment → paragraph verified Never state a higher level than was actually reached. ## KSC-specific traps ### Trap 1 — KSC is a Kosovo court, not an international tribunal The KSC is constitutionally and legally a **Kosovo court** (Article 162 of the Kosovo Constitution; Law 05/L-053). Calling it "an international tribunal" is imprecise. The KSC applies Kosovo law (informed by international law); its judges are international; but the institution itself is Kosovar. The accurate framing: **hybrid court / internationalised national court**. Not: "international tribunal". ### Trap 2 — "Statute" vs "Law" The KSC has **no separate Statute**. Law 05/L-053 is the equivalent of a Statute. Citing "the KSC Statute" is inaccurate; cite "the Law" or "Law 05/L-053". ### Trap 3 — Thaçi et al. is *awaiting judgment* As of early 2026, the Thaçi et al. case (KSC-BC-2020-06) is in the post-closing-statements phase, awaiting judgment. Hashim Thaçi has not been convicted of any crime. Care with framing — the accused are presumed innocent until conviction. The arrests occurred November 2020, indictment confirmed October 2020, trial began April 2023, closing statements February 2026. ### Trap 4 — Multiple Thaçi et al. cases There are now **two distinct Thaçi-headed cases**: - **KSC-BC-2020-06** — the principal war crimes / crimes against humanity case (Thaçi, Veseli, Selimi, Krasniqi) - **KSC-BC-2024-11** — an administration of justice case (Thaçi, Smakaj, Kilaj, Fazliu, Kuçi) — arrests December 2024, trial commenced February 2026 When the user refers to "the Thaçi case", clarify which. ### Trap 5 — Reparation orders are a substantive feature The KSC orders reparations directly against convicted persons (Law 05/L-053 Arts. 21-22; RPE Rule 168). The Shala Reparation Order of €208,000 (29 November 2024) is the leading example. This is distinct from the STL model (no reparations) and the ICC model (TFV-managed reparations). ### Trap 6 — Marty Report is parliamentary, not judicial The 7 January 2011 Marty Report is a **Council of Europe Parliamentary Assembly document**, not a judicial finding. When citing the Marty Report in the context of KSC jurisdiction (Law 05/L-053 narrows KSC competence to the allegations in the Report), note its status as a parliamentary inquiry document. ## Translation discipline KSC proceedings operate in **English, Albanian, and Serbian**. All public judgments and decisions are published in all three. The **English version is conventionally treated as authoritative for international scholarship**. Where the original speaker spoke Albanian or Serbian, the original-language version may matter for interpretation. The Law 05/L-053 exists in Albanian and Serbian as the original languages; English is an official translation. ## Partial verification is acceptable if disclosed If you can confirm the existence and substance but not the paragraph, the honest output says so. Disclosed partial verification is acceptable; silent overstatement is not. --- ## Reading the source document directly (the top of the ladder) The most reliable verification is reading the **actual document**, not a website's search snippet. Put this above everything else: **Rung 0 — work from the document itself when it is available.** Official tribunal sites frequently block automated fetching (HTTP 403), so a judgment can be public yet still unreachable by a direct fetch. Two ways to reach the text anyway: - **The user supplies it** — an uploaded PDF or pasted pages can be read directly, reaching paragraph-level verification. A practitioner working on a matter usually already holds the document; ask for it. - **A retrieval tool reads it** — where a document-retrieval tool or MCP server is available (one that fetches and extracts PDF text, with a fallback ladder), prefer it over a raw fetch. Only when the document cannot be obtained do you fall back to the search ladder above — and then you state the ceiling honestly. ## Site-search results are leads, not content A result from a site-search index — or a "synthesis" of search snippets — establishes at most that something **exists**. It is **never** content- or paragraph-level verification. Treat it as a lead to confirm against the document, and label it as such. Two recurring traps: - **Transliteration / OCR garbling.** Names and acronyms get corrupted (for example, an Arabic acronym surfacing as "KARA" where the source has "RADA/RADAA"). A name or acronym that appears only once in a snippet is a red flag — do not assert it. - **Relational claims.** Who is whose subordinate, associate, superior, or co-perpetrator is the detail a synthesis most often gets inverted. Never assert a relationship from a snippet; it requires the document.