arnautovd/vibecoding-claudecode-project-388

Purpose and Entry Points

This repository is a Claude-oriented automation workspace, not a conventional application. Its primary capability is a GitHub activity tracker that collects repository updates, detects changes, summarizes them, and sends Telegram notifications. A secondary skill extracts product prices from configured websites.

Primary operator instructions and behavior contracts are in:

• .claude/skills/tracker/SKILL.md

• .claude/skills/extract-price/SKILL.md

Tracker Orchestration

The tracker skill defines the workflow, target repositories, state expectations, and interaction with GitHub MCP/API tooling:

• .claude/skills/tracker/SKILL.md

• .claude/skills/tracker/targets.yaml

• .claude/skills/tracker/KNOWLEDGE.md

targets.yaml is the tracker’s source of truth for monitored entities. KNOWLEDGE.md supplies durable domain/context guidance for interpreting updates rather than merely forwarding raw events.

Collection and Change Detection

Collection obtains current GitHub data, while diffing compares snapshots or collected results to identify new/changed items:

• .claude/skills/tracker/scripts/collect.py

• .claude/skills/tracker/scripts/diff.py

• .claude/skills/tracker/references/github-mcp.md

The GitHub MCP integration is an external-service boundary. The rest of the pipeline should consume normalized collection output rather than directly coupling every stage to GitHub queries.

Transformation and Presentation

Detected updates are transformed into concise human-readable summaries and tabular views:

• .claude/skills/tracker/scripts/summarize.py

• .claude/skills/tracker/scripts/table.py

This forms the interpretation boundary: raw repository events become structured change records, then a notification-ready narrative or report. It separates collection/diff semantics from output formatting.

Delivery

Tracker output is delivered through a sending layer, including an explicit Telegram notifier:

• .claude/skills/tracker/scripts/send.py

• .claude/skills/tracker/scripts/notify_telegram.py

Telegram is the principal outbound external service. Delivery should remain downstream of collection, diffing, and summarization so notification failures do not alter source-state analysis.

Price Extraction Skill

The price-extraction skill runs a Python scraper/extractor against site-specific configuration:

• .claude/skills/extract-price/scripts/extract_price.py

• .claude/skills/extract-price/scripts/sites/www.regard.ru.yaml

• .claude/skills/extract-price/requirements.txt

The generic extraction runtime is separated from per-site selectors/rules in YAML. Its material boundary is the target retailer website; site changes should normally be handled in the site configuration before changing extraction logic.

Runtime and Automation

Python is the material execution runtime for both skills. Dependencies for the extraction workflow are declared separately in its requirements file. GitHub Actions provides repository validation/CI:

• .github/workflows/hexlet-check.yml

• .github/workflows/README.md

There is no evidenced deployed server, database, or web frontend. The repository is driven by skill instructions plus scripts, with GitHub/MCP as an input integration and Telegram as an output integration.