Agent skill

Hunk Diff Session Control

by modem-dev in modem-dev/hunk

“Interacts with live Hunk diff review sessions via CLI. Inspects review focus, navigates files, hunks, and exact lines, reloads session contents, adds inline review comments, and…”

— description from SKILL.md by modem-dev
MITAuto-check passedDevelopment

Install Hunk Diff Session Control

skills CLI
$ npx skills add modem-dev/hunk --skill hunk-review -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install modem-dev/hunk hunk-review --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/modem-dev/hunk.git skills-src && mkdir -p .claude/skills && cp -r skills-src/packages/hunk/skills/hunk-review .claude/skills/hunk-review && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
hunk-review
GitHub stars
9.5k
Token cost
~3.4k tokens
SKILL.md length
1,352 words
Files
1
Skills in repo
4
Repo updated
First seen
Licence
MIT

At a glance

  • Works in 5 steps: Load the right content (reload if needed) → Navigate to the first interesting file /… → Add a comment explaining what's… → …
  • SKILL.md covers Workflow, Session selection, Commands and New files in working-tree…, plus 2 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

About this skill

“Interacts with live Hunk diff review sessions via CLI. Inspects review focus, navigates files, hunks, and exact lines, reloads session contents, adds inline review comments, and paints attention marks on character ranges. Use when the user has a Hunk session running or wants to review diffs interactively.”

— description from SKILL.md by modem-dev

Hunk Diff Session Control is a skill in modem-dev/hunk (9.5k stars). Its SKILL.md is about 3.4k tokens. Licence: MIT.

Workflow steps

5 steps, taken from the first numbered list in SKILL.md.

  1. Load the right content (reload if needed)
  2. Navigate to the first interesting file / hunk
  3. Add a comment explaining what's happening and why
  4. If you already have several notes ready, prefer one comment apply batch over many separate shell invocations
  5. Summarize when done

What it can do on your machine

Read from SKILL.md and the folder at commit 252f59d. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md (its code samples are bash).

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Hunk Diff Session Control loads about 3.4k tokens when it runs. Until then it costs about 80 tokens; SKILL.md has 1,352 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~80
When it runs · the whole SKILL.md, loaded when a task matches
~3.4k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from modem-dev/hunk at commit 252f59d, republished under its MIT licence (© modem-dev). 1,352 words, ~3,426 tokens.

Download SKILL.mdSave it as .claude/skills/hunk-review/SKILL.md (or your agent's skills folder).
name
hunk-review
description
Interacts with live Hunk diff review sessions via CLI. Inspects review focus, navigates files, hunks, and exact lines, reloads session contents, adds inline review comments, and paints attention marks on character ranges. Use when the user has a Hunk session running or wants to review diffs interactively.

Hunk Review

Hunk is an interactive terminal diff viewer. The TUI is for the user -- do NOT run hunk diff, hunk show, or other interactive commands directly. Use hunk session * CLI commands to inspect and control live sessions through the local daemon.

If no session exists, ask the user to launch Hunk in their terminal first.

Workflow

text
1. hunk session list                                    # find live sessions
2. hunk session get --repo .                            # inspect path / repo / source
3. hunk session review --repo . --json                  # inspect file/hunk structure first
4. hunk session review --repo . --include-patch --json  # opt into raw diff text only when needed
5. hunk session context --repo .                        # check current focus when needed
6. hunk session navigate ...                            # move to the right place
7. hunk session reload -- <command>                     # swap contents if needed
8. hunk session comment add ...                         # leave one review note
9. hunk session comment apply ...                       # apply many agent notes in one stdin batch
10. hunk session highlight add ...                      # light up the exact range you are explaining

Session selection

Most session commands accept:

  • --repo <path> -- match the live session by its current loaded repo root (most common)
  • <session-id> -- match by exact ID (use when multiple sessions share a repo)
  • If only one session exists, it auto-resolves

reload also supports:

  • --session-path <path> -- match the live Hunk window by its current working directory
  • --source <path> -- load the replacement diff / show command from a different directory

Use --source only for advanced reloads where the live session you want to control is not already associated with the checkout you want to load next. For a normal worktree session, prefer selecting it directly with --repo /path/to/worktree.

Commands

Inspect
bash
hunk session list [--json]
hunk session get (<session-id> | --repo <path>) [--json]
hunk session context (<session-id> | --repo <path>) [--json]
hunk session review (<session-id> | --repo <path>) [--include-patch] [--include-notes] [--json]
  • get shows the session Path, Repo, and Source, which helps when choosing between --repo and --session-path
  • Repo is what --repo matches; Path is what --session-path matches
  • review --json returns file and hunk structure by default; add --include-patch only when a caller truly needs raw unified diff text
  • review --include-notes also returns the live review notes alongside the file and hunk structure
Navigate
bash
hunk session navigate (<session-id> | --repo <path>) --file <path> (--hunk <n> | --old-line <n> | --new-line <n>) [--json]
hunk session navigate (<session-id> | --repo <path>) --comment <id> [--json]
hunk session navigate (<session-id> | --repo <path>) (--next-comment | --prev-comment) [--json]

Absolute navigation requires --file and exactly one of --hunk, --new-line, or --old-line:

bash
hunk session navigate --repo . --file src/App.tsx --hunk 2
hunk session navigate --repo . --file src/App.tsx --new-line 372
hunk session navigate --repo . --file src/App.tsx --old-line 355

Exact comment navigation uses the commentId returned by hunk session comment list --json and does not require --file:

bash
hunk session navigate --repo . --comment comment-1

Relative comment navigation jumps between annotated hunks and does not require --file:

bash
hunk session navigate --repo . --next-comment
hunk session navigate --repo . --prev-comment
  • --hunk <n> is 1-based
  • --new-line / --old-line are 1-based line numbers on that diff side
  • A line target lands the user's viewport on that exact line (falling back to its hunk when the line is inside a collapsed region); --hunk lands on the hunk
  • Use either --next-comment or --prev-comment, not both
Reload

Swaps the live session's contents. Pass a Hunk review command after --:

bash
hunk session reload (<session-id> | --repo <path> | --session-path <path>) [--source <path>] [--json] -- diff [ref] [-- <pathspec...>]
hunk session reload (<session-id> | --repo <path> | --session-path <path>) [--source <path>] [--json] -- show [ref] [-- <pathspec...>]

Examples:

bash
hunk session reload --repo . -- diff
hunk session reload --repo . -- diff main...feature -- src/ui
hunk session reload --repo . -- show HEAD~1
hunk session reload --repo . -- show HEAD~1 -- README.md
hunk session reload --repo /path/to/worktree -- diff
hunk session reload --session-path /path/to/live-window --source /path/to/other-checkout -- diff
  • Always include -- before the nested Hunk command
  • --repo or <session-id> usually selects the session you want
  • --source is advanced: it does not select the session; it only changes where the replacement review command runs
  • If the live session is already showing the target worktree, prefer hunk session reload --repo /path/to/worktree -- diff
  • --session-path targets the live window when you need to keep session selection separate from reload source
Comments
bash
hunk session comment add (<session-id> | --repo <path>) (--reply-to <note-id> | --file <path> (--old-line <n> | --new-line <n>)) --summary <text> [--rationale <text>] [--author <name>] [--markup <stml>] [--focus] [--json]
hunk session comment apply (<session-id> | --repo <path>) --stdin [--focus] [--json]
hunk session comment list (<session-id> | --repo <path>) [--file <path>] [--type <live|all|ai|agent|user>] [--json]
hunk session comment rm (<session-id> | --repo <path>) <comment-id> [--json]
hunk session comment clear (<session-id> | --repo <path>) [--file <path>] [--include-user|--all] --yes [--json]

Examples:

bash
hunk session comment add --repo . --file README.md --new-line 103 --summary "Tighten this wording"
hunk session comment add --repo . --reply-to user:123 --summary "Addressed in the latest revision"
printf '%s\n' '{"comments":[{"filePath":"README.md","newLine":103,"summary":"Tighten this wording"}]}' | hunk session comment apply --repo . --stdin
  • comment list --type user shows human-authored inline notes; without --type, comment list preserves the legacy live-agent-comment view
  • comment add is best for one note; comment apply is best when an agent already has several notes ready
  • Root comment add notes require --file, --summary, and exactly one of --old-line or --new-line; replies use --reply-to <note-id> with --summary and inherit the parent's anchor
  • comment apply items require summary plus either replyTo by itself or filePath with exactly one target such as hunk, hunkNumber, oldLine, or newLine
  • comment apply reads a JSON batch from stdin and validates the full batch before mutating the live session
  • Pass --focus when you want to jump to the new note or the first note in a batch
  • comment list and comment clear accept optional --file
  • Quote --summary and --rationale defensively in the shell
Attention marks

Highlights paint character ranges inside the diff lines the user is looking at — use them to light up the exact expression you are explaining while you narrate.

bash
hunk session highlight add (<session-id> | --repo <path>) --file <path> (--old-line <n> | --new-line <n>) --start <n> --end <n> [--tone <tone>] [--focus] [--json]
hunk session highlight clear (<session-id> | --repo <path>) [--file <path>] [--json]

Examples:

bash
hunk session highlight add --repo . --file src/App.tsx --new-line 42 --start 6 --end 19
hunk session highlight add --repo . --file src/App.tsx --new-line 42 --start 6 --end 19 --tone warning --focus
hunk session highlight clear --repo .
  • highlight add requires --file, exactly one of --old-line or --new-line, and the --start / --end offsets
  • --start is a 0-based inclusive offset into the line's text and --end is exclusive, counted in UTF-16 code units — the same [start, end) range extensions use
  • Tones: match (default), info, warning, error, dim; current renders as reverse video and is best reserved for the one range under discussion
  • Pass --focus to also land the viewport on the marked line
  • Marks survive scrolling, navigation, and reloads that leave the marked file's content unchanged; a reload that changes that file drops its marks, and highlight clear removes them explicitly (optionally per --file)
  • Marks are visual only — pair them with a comment add when the explanation should persist as a note
Experimental rich markup notes (STML)

Only use STML when hunk session context --json lists stml in experimentalFeatures. The user opts into that experience by launching the review with --experimental; do not ask a normal session to render markup.

For an opted-in session, --markup (or a markup field on apply items) renders the note body as STML — a small HTML-like markup for terminal UI (boxes, rows, gauges, badges, lists, code). Keep --summary a real sentence: it is the fallback and the comment list text.

Before writing markup, run hunk markup guide once — it has copy-paste patterns and the width rules. The session context also reports noteMarkupWidth (the live render width); preview with hunk markup render - --width <that>. Comment responses echo markupWidth and return markupNotes when markup degraded — fix what they flag.

Show full SKILL.md (534 more words)Show less

New files in working-tree reviews

hunk diff includes untracked files by default. If the user wants tracked changes only, reload with --exclude-untracked:

bash
hunk session reload --repo . -- diff --exclude-untracked

Guiding a review

The user may ask you to walk them through a changeset or review code using Hunk. Start with hunk session review --json to understand the file/hunk structure without inflating agent context, then use --include-patch only for the files you truly need to read in raw diff form. Use context and navigate to line up the user's current view before adding comments.

Your role is to narrate: steer the user's view to what matters and leave comments that explain what they're looking at.

Typical flow:

  1. Load the right content (reload if needed)
  2. Navigate to the first interesting file / hunk
  3. Add a comment explaining what's happening and why
  4. If you already have several notes ready, prefer one comment apply batch over many separate shell invocations
  5. Summarize when done

Guidelines:

  • Work in the order that tells the clearest story, not necessarily file order
  • Navigate before commenting so the user sees the code you're discussing
  • Use highlight add --focus to steer the user's eyes to the exact expression while you explain it, and highlight clear before moving to the next topic
  • Use comment apply for agent-generated batches and comment add for one-off notes
  • Use --focus sparingly when the note itself should actively steer the review
  • Keep comments focused: intent, structure, risks, or follow-ups
  • Don't comment on every hunk -- highlight what the user wouldn't spot themselves

Common errors

  • "No diff file matches ..." -- the file is not in the loaded review. Check context, then reload if needed.
  • "No active Hunk sessions" -- if Hunk is visibly running, localhost may be blocked by the agent sandbox; retry with network/sandbox escalation. Otherwise ask the user to open Hunk.
  • "Multiple active sessions match" -- pass <session-id> explicitly.
  • "No active session matches session path ..." -- for advanced split-path reloads, verify the live window Path via hunk session get or list, then use --session-path.
  • "Pass the replacement Hunk command after --" -- include -- before the nested diff / show command.
  • "Pass --stdin to read batch comments from stdin JSON." -- comment apply only reads its batch payload from stdin.
  • "Specify exactly one navigation target" -- pick one of --hunk, --old-line, or --new-line.
  • "Specify exactly one comment target" -- pass comment add one of --old-line or --new-line.
  • "Specify exactly one highlight target" -- pass highlight add one of --old-line or --new-line.
  • "Highlight --end must be greater than --start" -- offsets are [start, end) UTF-16 code units into the line text; end is exclusive.
  • "Specify either --next-comment or --prev-comment, not both." -- choose one comment-navigation direction.
  • "The session daemon is ..." -- a daemon-build-mismatch (the --json error carries daemon, cli, attachedSessions, and recommendedAction). Tell the user which build is newer and how many windows are attached, then ask before running hunk daemon restart --yes; never restart unprompted. After the restart, windows that failed to register attach on their own, so re-run hunk session list instead of relaunching anything. When recommendedAction is use-newer-hunk, the daemon is the newer build: use that Hunk instead.
  • "Could not read the raw diff for ..." -- the session reloaded or closed while --include-patch was reading it. Re-run review; drop --include-patch if you only need file and hunk structure.

© modem-dev, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in packages/hunk/skills/hunk-review of modem-dev/hunk.

Open the folder on GitHubat commit 252f59d

Compare with similar skills

Hunk Diff Session Control next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Hunk Diff Session Control compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Hunk Diff Session Control this skillmodem-dev/hunk9.5k—~3.4kAutomated safety check: PassMIT
Code Review ChecklistshareAI-lab/learn-claude-code78k5 repos~1.1kAutomated safety check: PassMIT
Understand Diff AnalysisEgonex-AI/Understand-Anything85k1 repos~1.4kAutomated safety check: PassMIT
Open Code Review CLIalibaba/open-code-review44k—~3.1kAutomated safety check: PassApache-2.0
Open Code Review Delegatealibaba/open-code-review44k—~2kAutomated safety check: PassApache-2.0
PR Review State Fetchprisma/orm48k—~767Automated safety check: PassApache-2.0

Similar skills

  • Code Review Checklist

    shareAI-lab/learn-claude-code

    Reviews code against a five-part checklist covering security, correctness, performance, maintainability and testing, and reports findings in a fixed format.

    78k GitHub starsUsed in 5 repos~1.1k tokens
    DevelopmentAuto-check passed
  • Understand Diff Analysis

    Egonex-AI/Understand-Anything

    Reads your git changes or a pull request against a prebuilt knowledge graph of the project to explain what changed, which components are affected and what is risky.

    85k GitHub starsUsed in 1 repo~1.4k tokens
    DevelopmentAuto-check passed
  • Open Code Review CLI

    alibaba/open-code-review

    Runs the ocr command-line tool to review Git changes, a commit or a branch comparison with an AI model, returning line-level comments and optionally applying fixes.

    44k GitHub stars~3.1k tokensUpdated 2 days ago
    DevelopmentAuto-check passed
  • Open Code Review Delegate

    alibaba/open-code-review

    Has the host agent do the code review itself while the ocr CLI handles file selection and rule lookup, covering workspace changes, branch ranges or single commits.

    44k GitHub stars~2k tokensUpdated 2 days ago
    DevelopmentAuto-check passed
  • Official

    Fetches a pull request's canonical review state as JSON, validates it, and renders markdown, a text summary and triage target files from it using bundled scripts.

    48k GitHub stars~767 tokensUpdated today
    DevelopmentAuto-check passed
  • Knowledge Graph PR Review

    tirth8205/code-review-graph

    Reviews a pull request or branch diff with a code knowledge graph and produces a structured review that includes blast-radius analysis.

    32k GitHub stars~452 tokensUpdated today
    DevelopmentAuto-check passed

More from modem-dev/hunk

  • Hunk Release Workflow

    modem-dev/hunk

    Maintainer workflow for preparing, publishing, verifying and curating Hunk releases, with confirmation gates before tags, publishes and public edits.

    9.5k GitHub stars~3.8k tokensUpdated today
    Auto-check passed
  • Maps the hunkdiff/extension API for building Hunk terminal diff viewer extensions: panes, file views, commands, dialogs, themes, VCS backends and lifecycle events.

    9.5k GitHub stars~5.8k tokensUpdated today
    Auto-check passed
  • Produces captioned 1080p product videos of the Hunk terminal diff viewer by driving the real TUI headlessly and compositing frames in Chromium.

    9.5k GitHub stars~5.5k tokensUpdated today
    Auto-check passed

Works with

Categories

Questions about Hunk Diff Session Control

How do I install Hunk Diff Session Control in Claude Code?

Run `npx skills add modem-dev/hunk --skill hunk-review -a claude-code`. Or copy the skill folder (packages/hunk/skills/hunk-review in modem-dev/hunk) into .claude/skills/hunk-review in your project. Claude Code loads it when a task matches its description.

How do I install Hunk Diff Session Control in Codex?

Run `npx skills add modem-dev/hunk --skill hunk-review -a codex`. Or copy the skill folder (packages/hunk/skills/hunk-review in modem-dev/hunk) into .agents/skills/hunk-review in your project. Codex loads it when a task matches its description.

Can I use Hunk Diff Session Control in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add modem-dev/hunk --skill hunk-review -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/hunk-review, .gemini/skills/hunk-review, .github/skills/hunk-review and .opencode/skills/hunk-review in your project.

What does Hunk Diff Session Control need to run?

SKILL.md names no scripts, command-line tools or credentials: Hunk Diff Session Control is instructions for the agent only.

Does Hunk Diff Session Control access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Hunk Diff Session Control safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Hunk Diff Session Control use?

Hunk Diff Session Control is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Hunk Diff Session Control use?

About 3.4k tokens (SKILL.md is roughly 14k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Hunk Diff Session Control?

Skills that share tags, products or a category with Hunk Diff Session Control: Code Review Checklist (shareAI-lab/learn-claude-code, 78k stars), Understand Diff Analysis (Egonex-AI/Understand-Anything, 85k stars), Open Code Review CLI (alibaba/open-code-review, 44k stars) and Open Code Review Delegate (alibaba/open-code-review, 44k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Hunk Diff Session Control?

modem-dev (a GitHub organization) maintains it in modem-dev/hunk, which has 9,521 GitHub stars. The repository holds 4 skills in this directory. The repository was last updated on October 6, 2026.

Source: modem-dev/hunk on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.