intelscan 0.1.3

Workspace summary generator for AI-assisted development workflows.

IntelScan

IntelScan is a lightweight workspace-summary tool for AI-assisted development workflows. https://pypi.org/project/intelscan

It scans a repository, writes a structured manifest for tools, and generates a human-readable memory file so future agent passes can understand the project without rescanning everything from scratch.

By default, scans skip common large/generated folders such as .git, node_modules, .venv, and __pycache__. IntelScan also loads .intelscanignore and respects .gitignore when present.

What it generates

• workspace.json - machine-readable workspace metadata
• workspacememory.md - human-readable workspace summary

These files are generated at the workspace root and are intended to be refreshed as the project changes.

Why it helps

• Gives agents a fast starting point before broad repo scans
• Captures file inventory, package metadata, project structure, and Git status when available
• Keeps a stable handoff artifact for repeated AI-assisted workflows
• Supports both one-off scans and automated refresh flows

Installation

Prerequisites:

• Python 3.8+
• Git on PATH if you want Git metadata included in the generated summary

Install from PyPI:

pip install intelscan

Install a specific release:

pip install intelscan==0.1.1

Install from source:

git clone https://github.com/Debanshu2005/IntelScan
cd IntelScan
python -m pip install .

Quick Start

Run a one-time scan:

intelscan --root .

Run in watch mode:

intelscan --root . --watch

Create a project-local agent guide plus companion agent instruction files when they do not already exist:

intelscan --root . --init-agents

Use your AI agent's CLI mode via IntelScan, with automatic manifest refresh before and after the pass:

intelscan-agent --root . --agent-cmd "codex"
intelscan-agent --root . --agent-cmd "claude"
intelscan-agent --root . --agent-cmd "myagent --help"

The coordinator also exports INTELSCAN_WORKSPACE_ROOT, INTELSCAN_WORKSPACE_JSON, and INTELSCAN_WORKSPACE_MD for agent wrappers that want to consume the generated context automatically. On Windows, intelscan-agent first tries normal PATH lookup, can search additional directories with --agent-search-dir, and falls back to VS Code extension bin folders when the requested agent CLI is bundled there.

For source-tree development without installing globally:

python workspace_scanner.py --root .
python workspace_scanner.py --root . --watch
python agent_coordinator.py --root . --agent-cmd "python your_agent_task.py"

Agent Guide Bootstrap

--init-agents creates AGENTS.md as the canonical shared guide and also bootstraps companion files for common agent ecosystems when missing:

• CLAUDE.md
• GEMINI.md
• .github/copilot-instructions.md

Existing files are preserved and are never overwritten.

Main Commands

Scanner CLI:

intelscan --root .
intelscan --root . --watch
intelscan --root . --output-json workspace.json --output-md workspacememory.md
intelscan --root . --exclude-dir .cache
intelscan --root . --max-depth 4 --progress

Coordinator CLI:

intelscan-agent --root . --agent-cmd "codex"
intelscan-agent --root . --agent-cmd "claude" --always-refresh
intelscan-agent --root . --agent-cmd "myagent --help" --skip-initial-scan
intelscan-agent --root . --agent-cmd "myagent --help" --agent-search-dir "D:\Tools\Agents"

Repository Layout

• src/intelscan/workspace_scanner.py - scanner implementation and CLI entry point
• src/intelscan/agent_coordinator.py - agent wrapper that refreshes workspace context around a command
• workspace_scanner.py - thin source-tree launcher for local development
• agent_coordinator.py - thin source-tree launcher for local development
• WORKSPACE_SCANNER.md - focused scanner behavior and usage notes
• tests/test_workspace_scanner.py - coverage for manifest and safety behavior

The project uses a src/intelscan/ package layout. The root Python files are intentionally thin launchers so local development commands still work directly from the repository.

Development

Run tests:

python -m unittest -q

Compile-check the main scripts:

python -m py_compile workspace_scanner.py agent_coordinator.py src/intelscan/workspace_scanner.py src/intelscan/agent_coordinator.py

Build a wheel locally when verifying packaging behavior:

python -m pip wheel --no-build-isolation --no-deps --wheel-dir .tmp_wheelhouse .

Notes

• Output writes are constrained to the workspace root.
• Symlinked files and directories are not followed during scans.
• Git metadata is collected with safe non-interactive commands when Git is available.
• Use .intelscanignore for scanner-specific exclusions, or --no-gitignore if you need to ignore .gitignore during a scan.
• Use --max-depth and --progress when scanning very large repositories or monorepos.

Publishing

Release guidance lives in RELEASING.md. The repository also includes a GitHub Actions workflow for PyPI Trusted Publishing at .github/workflows/publish.yml.