Legal Matter Tracker

Legal Matter Tracker

This skill scans a local Cowork workspace for files related to a specified client or legal matter, extracts dated events and key facts, and assembles a structured chronological timeline report — entirely from local files, with no external integrations required.

Input: Client or case name (required). Optional: folder path to narrow scan scope.

Output: Structured markdown timeline report (inline). Optionally saved to output/matter-timeline-[client]-YYYY-MM-DD.md on request.

---

Language Detection

Detect the user's language from their message:

• If Russian (or contains Cyrillic): respond in Russian
• If English (or other Latin-script language): respond in English
• If ambiguous: default to English

---

Instructions

Step 1: Parse Input

1. Extract the client or case name from the user's message.

• Accepted forms: "case timeline for Acme Corp", "track Smith matter", "хронология по делу Иванов", "что было по клиенту Ромашка".
• The name may be a company, person, project code, or case identifier.

1. If no name is provided, return:

• EN: "Please provide a client or case name. Example: 'Case timeline for Acme Corp'"
• RU: "Укажите имя клиента или название дела. Пример: «хронология по делу Ромашка»"

Stop — do not generate output.

1. If the user provided a folder path to narrow the scan scope, note it. Otherwise scan the full workspace.

---

Step 2: Scan Workspace

1. Search for files that match the client/case name using these criteria:

• Filename match: filename contains any word from the name (case-insensitive).
• Content match: file contains the name in a heading (#, ##, **Name**) or first 10 lines.
• Accepted file types: .md, .txt.
• Search folders: workspace root + notes/, clients/, cases/, input/, output/, context/.

1. If a folder path was provided: scan only that folder (recursive).

1. Count total files scanned and matching files found.

1. If no matching files found, return:

• EN: "No files found matching '[name]'. Try a broader term or check that the relevant files are in your workspace."
• RU: "Файлов по запросу «[name]» не найдено. Попробуйте более широкий запрос или проверьте, что нужные файлы есть в воркспейсе."

Stop — do not generate output.

---

Step 3: Extract Events

For each matched file:

1. Extract dated events: lines or paragraphs that contain:

• An explicit date (ISO: 2025-01-15, written: January 15, 2025, 15 января 2025, 15.01.2025, 01/15/2025)
• A fact, decision, event, action item, or key information associated with that date.

1. Extract key entities (if detectable):

• Party names (people, organisations)
• Amounts or values (monetary, percentages, quantities)
• Status indicators (open, closed, pending, active, закрыто, открыто)
• Action items or open questions

1. For lines with no associated date: collect as undated entries — do not discard.

1. If a file yields zero events (file is empty or contains only unstructured text with no extractable facts), note it and skip.

1. If all matched files yield zero events, return:

• EN: "Found [N] file(s) matching '[name]' but could not extract events. Files may be empty or contain only unstructured text."
• RU: "Найдено [N] файл(ов) по запросу «[name]», но извлечь события не удалось. Файлы могут быть пустыми или содержать только неструктурированный текст."

Stop.

---

Step 4: Assemble Timeline

1. Sort all dated events chronologically, ascending (oldest first).

1. Group undated entries separately at the end — ordered by source file.

1. Build the timeline table:

• Column 1: Date (formatted as YYYY-MM-DD for ISO dates; preserve written format for ambiguous dates)
• Column 2: Event (concise description, max 120 chars; truncate with … if needed)
• Column 3: Source (filename with relative path, in backticks)

1. Build the Key Facts Summary using extracted entities:

• Parties (if found)
• Matter status (open/closed/unknown)
• Open items (action items or questions found in files)

1. Build the Sources list: each matched file with count of extracted events.

1. Assemble the full report (see Output Format below).

---

Step 5: Output

1. Display the complete timeline report inline in chat.

1. If the user says "save" or "save to file" (or RU: "сохрани", "сохрани в файл"):

• Write to output/matter-timeline-[client-slug]-YYYY-MM-DD.md where client-slug is the client name lowercased with spaces replaced by hyphens.
• Confirm: "Saved to output/matter-timeline-[client-slug]-YYYY-MM-DD.md"

---

Edge Cases

• Multiple files match: Include all matched files; label every event with its source filename. Do not deduplicate events.
• No explicit dates in matched files: Include all entries in the Undated Entries section. Add a note at the top: "No dated events found — all entries are undated."
• User provides folder path to narrow scan: Scan only that folder (recursive); note scan scope in report header.
• Conflicting facts across files (e.g., different amounts for the same event): List both versions in the event table with their respective source references. Do not resolve silently; add a flag ⚠ Conflict in the Event column.
• Very long event descriptions (>120 chars): Truncate with … in the table; include the full text as a footnote below the table.
• Date formats in Russian (e.g., 15 января 2025): Parse and normalise to YYYY-MM-DD for chronological sorting; display original format in the table.

---

Negative Cases

• No client/case name provided: Return usage message (EN/RU). Stop. No partial output.
• No matching files found: Return "No files found" message with search guidance. Stop.
• Files found but no extractable events: Return informative message. Stop. Do not generate an empty timeline.

---

Output Format

# Matter Timeline — [Client/Case Name]
**Generated:** YYYY-MM-DD
**Files scanned:** N  |  **Matching files:** M  |  **Events extracted:** K
**Scan scope:** [workspace root / specified folder path]

---

## Chronological Timeline

| Date | Event | Source |
|------|-------|--------|
| 2025-01-15 | Initial consultation. Client presented IP dispute claim. | `notes/acme-notes.md` |
| 2025-02-03 | Contract signed. Engagement fee: $12,000. | `contracts/acme-contract.md` |
| 2025-03-10 | Discovery phase completed. 3 key documents identified. | `notes/acme-notes.md` |

---

## Undated Entries
*(entries without extractable dates — ordered by source file)*

| Entry | Source |
|-------|--------|
| Opposing counsel: Smith & Partners LLP | `notes/acme-notes.md` |

---

## Key Facts Summary
- **Parties:** Acme Corp (client), Smith & Partners LLP (opposing counsel)
- **Matter status:** Open (last activity: 2025-03-10)
- **Open items:**
  - Discovery documents pending review
  - Fee invoice not yet confirmed

---

## Sources
- `notes/acme-notes.md` — 3 events extracted
- `contracts/acme-contract.md` — 1 event extracted

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: KirKruglov
• Source: KirKruglov/claude-skills-kit
• License: MIT

Install and usage instructions live in the source repository linked above.