原始内容
Social value
Produce social value data from imported data for our social value report for Islington Council.
Architecture
Runtime Environment
This application runs within the AI coding interface:
- Application runs via natural language interaction with AI assistant
- AI assistant is the interface
- No API keys or separate deployment needed
- Skills provide capabilities
- Simple file-based storage (JSON, CSV, etc.)
Technical Stack
- Language: Python
- Documentation: British English
- Commits: Conventional Commits specification
Project Structure
project/
├── AGENTS.md # Project conventions and session bootstrap
├── README.md # This file
├── CLAUDE.md # Claude Code auto-load redirect (optional)
├── skills/ # Reusable AI capabilities
│ ├── spec.md # Creates project specifications
│ └── commit.md # Formats commit messages
└── [project files] # Your application code
Getting Started
Session Bootstrap
To start working with this project in any AI coding assistant:
Codex CLI: AGENTS.md loads automatically - just start working Claude Code: Say anything (if CLAUDE.md is set up), or "Read AGENTS.md" Other tools: Say "Read AGENTS.md"
Available Skills
Current skills in skills/ directory:
- spec - Create project specifications
- commit - Format commit messages following Conventional Commits
- refactor - Analyse Python code for refactoring opportunities and quality issues
- social-value-report - Generate the social value report from anonymised CSV data
Global skills in ~/.skills/ directory:
- anonymise - Strip PII from CSVs and timestamp outputs
Configuration files
~/.skills/anonymise/config.txt– column headers to remove during anonymisation (one per line,#for comments)skills/social-value-report/config.json– mapping from report fields to CSV column headers (defaults prefer aim 2, fall back to aim 1)
To add more skills, create skills/skillname.md with YAML frontmatter and instructions.
Development Workflow
Spec-First Development
- Start with problem statement
- Create specification (what the system does)
- Define user stories (features from user perspective)
- Implement iteratively, one story at a time
- Use conventional commits for clean history
Creating Skills
Skills are markdown files with YAML frontmatter:
---
name: skill-name
description: When to use this skill. AI uses this to auto-activate.
---
# Skill instructions here
Simple skills go in skills/skillname.md. Complex skills with supporting files go in skills/skillname/SKILL.md.
Platform Agnostic
This project works with:
- Claude Code
- Codex CLI (OpenAI)
- Gemini CLI (Google)
- Cursor
- Aider
- GitHub Copilot CLI
- Any future AI coding assistant
Skills and AGENTS.md are plain markdown - no vendor lock-in.
References
For detailed conventions, coding standards, and session bootstrap process, see AGENTS.md.