Anna AI-Native 2026 · Developer Tools

ShipGhost

Twenty messy commits → one clean PR. Your friendly release-engineering ghost.

Anna App · schema 2 AES-GCM-256 GPG clearsign

✦

Slide 02 // The Problem

PR knowledge debt

After a long sprint your git log is twenty commits of wip, fix, stuff. Writing clean conventional commits and a real PR summary is tedious — so most devs skip it.

Lost context

Reviewers can't tell what changed or why.

Tedious busywork

Summaries + conventional commits eat the end of every sprint.

AI helpers leak code

They ship raw, proprietary diffs in plaintext to cloud LLMs.

Slide 03 // The Solution

A local AI release engineer

①Reads your branch with real git diff/git log — locally.
②Ghostwrites the PR — title, summary, rationale, inline comments — and cleans commit messages.
③You review & sign — GPG/SSH clearsign proves a human approved it.

Diffs are AES-GCM-256 enveloped before the LLM ever sees them — source IP stays on your machine.

Slide 04 // Core Product Flow

Analyze → Ghostwrite → Clean → Sign

git.analyze

diff + log subprocess

➜

pr.ghostwrite

LLM PR package

➜

commits.cleanup

conventional commits

➜

pr.sign

GPG clearsign → R2

Draft history persisted to Anna APS KV (last 50 PRs).

Slide 05 // Technical Architecture

Secure, native Anna app

App UI (iframe)

3 views: main · inline_inspector · commit_cleaner
tools.invoke → real Executa calls
chat.append_artifact PR card
upload → signed bundle on R2

Executa (Python, stdio JSON-RPC)

Real git subprocess analysis
Reverse-RPC sampling/createMessage
storage APS KV draft history
AES-GCM-256 + GPG/SSH clearsign

anna-app validate --strict passes · boots in anna-app dev · 107 tests passing.

Slide 06 // Live Demo Highlights

What the judges see

1

Point at a repo

Real git diff against the base branch.

2

PR draft appears

Title, summary, rationale, inline comments.

3

Commit cleaner

Messy → conventional, side by side.

4

Clearsigned export

GPG-signed markdown bundle to R2.

No Anna account? It also runs standalone in a browser with a sample diff.

Slide 07 // Fit with Anna

We use the platform, deeply & for real

Executa stdio tools sampling/createMessage storage (APS KV) upload → R2 chat.append_artifact 3-view multi-window

Plus a sandboxed Developer Console that exercises the Host-API surface live — routing to the real platform inside Anna, returning clearly-labeled mock responses standalone. Transparent by design.

Slide 08 // Who needs this

Every developer opens PRs

100M+

developers — billions of pull requests a year.

DevEx

PR quality & review speed are core team metrics.

Private

Enterprises that can't send source diffs to cloud AI.

Slide 09 // Competitive Edge

Why ShipGhost wins

VS

writing it by hand — seconds, not the dreaded end-of-sprint hour.

VS

cloud AI tools — AES-GCM-256 diff envelopes + local GPG signing; IP never leaves in plaintext.

VS

PR templates — it actually reads the diff and explains the architecture impact.

And it's a joy. A cute ghost that does your worst chore — devs want to ship with it.

Slide 10 // Roadmap

30 / 60 / 90

30 DAYS

Git hooks

Auto-run ShipGhost on push.

60 DAYS

Multi-branch diffs

Compare staging ↔ production.

90 DAYS

Durable PR archive

Browse & re-download every PR (APS Files).

Slide 11 // Team

Built solo, shipped clean

107 plugin tests passing
Real git subprocess (no mocks in core)
anna-app validate --strict green
Original animated ghost mascot & brand

Ship it. Let the ghost write.

ShipGhost kills PR knowledge debt — securely, joyfully, and natively on Anna.

Useful Working demo Real AI Native to Anna Delightful