Anchored Agentic Development

Anchored Agentic Development

*Against Entropy*

Resources for the AAD methodology — a framework for keeping agentic development coherent over time.

→ [Full framework](https://nbaglivo.dev/aad)

---

What's here

**`/templates`** — ADR template. Copy it, fill it in, commit it. The template is the starting point for every architectural decision you capture.

**`/skills/adr`** — Claude Code skill for creating ADRs. Enforces structure, checks for conflicts against existing ADRs, generates the file. Use with `/adr`.

**`/skills/review-adr`** — Claude Code skill for validating a draft ADR before it goes to peer review. Checks for missing rationale, conflicts with prior decisions, and superseded decisions. Use with `/review-adr`.

---

The short version

Agents and developers are building from your codebase continuously. Without shared architectural context, that work drifts — each agent making its own decisions, patterns diverging, the system fragmenting quietly.

AAD is about setting the anchor before the work starts. You capture architectural decisions as ADRs. Those ADRs become the context that any agent, any developer, any automated workflow can draw from. The architecture stays coherent not because someone is reviewing everything, but because the direction was there before the build began.

Read the full framework for the complete picture.

---

Using the skills

Install a skill by adding it to your Claude Code project. Then call it by name during a session.

`/adr` — starts the ADR creation process. Loads your existing ADR corpus, helps you think through the decision, writes the draft, and checks for conflicts.

`/review-adr` — validates a draft ADR. Point it at a file and it checks completeness, conflict with prior decisions, and anything that should be named before the PR goes up.

---

Contributing

If you build something that fits this methodology — a skill, a tool, an integration — open a PR or reach out.

[nicolas.baglivo@gmail.com](mailto:nico