git-hunk

The problem

Partial staging is interactive-only

The only built-in way to stage individual hunks is git add -p — an interactive prompt that blocks automation.

Interactive staging

$ git add -p
@@ -42,6 +42,8 @@
+    if (flags.verbose) {
+        log.info("enabled");
+    }

(1/3) Stage this hunk [y,n,q,a,d,e,?]? █

✗ Requires a human at the keyboard
✗ Blocks CI/CD pipelines
✗ Non-deterministic hunk ordering
✗ Unusable by LLM agents

Hash-based staging

$ git hunk list --oneline
a3f7c21  src/main.zig   42-49  if (flags.verbose) {…
b82e0f4  src/parse.zig  15-28  fn parseArgs(alloc: …
c91d3a8  src/args.zig   8-8    const quiet: bool = …

$ git hunk add a3f7c21
staged a3f7c21 → a3f7c21  src/main.zig

✓ Fully non-interactive
✓ Stable SHA-1 content hashes
✓ Deterministic across runs
✓ Machine-readable output

How it works

Enumerate. Select. Stage.

Three commands. No prompts. No interaction.

1

Enumerate

$ git hunk list

See every hunk with a stable content hash. Same content always produces the same hash.

2

Select

$ git hunk diff a3f7

Inspect any hunk by its hash prefix. Review the full diff before staging.

3

Stage

$ git hunk add a3f7

Stage by hash — deterministic and scriptable. Remaining hunk hashes stay stable.

The key insight

Hashes that don't break

When you stage a hunk, the remaining hashes stay the same. This is what makes scripted multi-step workflows reliable.

Before staging

a3f7c21  src/main.zig   42-49  if (flags.verbose) {…
b82e0f4  src/parse.zig  15-28  fn parseArgs(alloc: …
c91d3a8  src/args.zig   8-8    const quiet: bool = …


$ git hunk add a3f7c21

After staging

b82e0f4  src/parse.zig  15-28  fn parseArgs(alloc: …  ← same hash
c91d3a8  src/args.zig   8-8    const quiet: bool = …  ← same hash

How it works


SHA1( file_path + '\0' + stable_line + '\0' + diff_lines )

Hashes use the immutable side's line numbers. Staging one hunk can't shift another hunk's anchor point, so hashes remain stable across operations.

Built for

Stage hunks by hash, not by prompt

>_

LLM Agents

Agents can't drive interactive prompts. git-hunk gives them deterministic hashes to enumerate, select, and stage programmatically.

--porcelain

#!/

Scripts & CI

Deterministic staging in pipelines. Use count as a guard and check to validate hashes before staging.

--exclusive

$

Humans

No more y/n/q/a/d/e/? guessing. See all hunks at once, inspect any hunk by hash, stage exactly what you want.

--oneline

Commands

Everything you need

list

Enumerate hunks with stable content hashes

Compact output

git hunk list --oneline

View staged hunks

git hunk list --staged

Machine-readable

git hunk list --porcelain

diff

Inspect a hunk's full diff before staging

View by hash prefix

git hunk diff a3f7

Specific lines

git hunk diff a3f7:3-5

count

Count available hunks — bare integer for scripting

CI guard

git hunk count

Install

Get started

Homebrew

brew install shhac/tap/git-hunk

AI Agent Skill

npx skills add shhac/git-hunk

GitHub Release (macOS)

curl -L https://github.com/shhac/git-hunk/releases/latest/download/git-hunk-macos-universal.tar.gz | tar xz

Build from Source

git clone https://github.com/shhac/git-hunk.git && cd git-hunk && zig build -Doptimize=ReleaseFast

Demo