Project scope

Kaddo is an open-source CLI and agent prompt toolkit for building a living knowledge layer close to the code. This page makes its current scope explicit.

What Kaddo does

• Initializes a project knowledge structure (kaddo init).
• Detects deterministic repo signals (kaddo scan).
• Creates LLM context packs (kaddo context).
• Installs agent prompt packs (kaddo add agents).
• Guides the CLI → LLM handoff (kaddo understand).
• Creates Work Items from a roadmap (kaddo create --from roadmap).
• Helps declare code ownership (kaddo owners suggest).
• Detects possible knowledge drift (kaddo guard).
• Explains the project state (kaddo explain).

What Kaddo does not do

• It does not call an LLM by default.
• It does not require an API key.
• It does not generate code.
• It does not replace human review.
• It does not infer business truth automatically.
• It does not replace Jira, Linear or GitHub Issues.
• It does not understand legacy systems magically.

The two layers

Layer	Responsibility
CLI (deterministic)	scan signals, package context, install agent prompts, guide handoff, create Work Items, declare ownership, detect drift, explain state
LLM (interpretation)	extract capabilities, reconstruct architecture, propose a roadmap, identify risks, draft structured artifacts

The CLI prepares and stores context; your LLM interprets it using Kaddo agents.

Supported project states

• new — start with a minimal knowledge structure from day one.
• pre-AI — prepare an existing repo for humans and LLM agents.
• legacy — understand and reduce risk before changing fragile systems.

See the per-state guides: New project, Pre-AI project, Legacy project.

Current limitations

• The interpretation steps (capabilities, architecture, roadmap, legacy analysis) happen in your LLM chat using Kaddo agents — they are not produced by the CLI.
• Drift detection is glob-based (code: ownership vs git diff); it does not perform semantic code analysis.
• Ownership assistance currently targets Work Items.