ClaudeCodeMod

All shelves / Mods / Review

assumption-check

By vumichien

assumption-check is a Claude Code review mod by vumichien.

Lists the choices Claude made that you did not state after a turn that edited files or ran commands; mark each right or wrong and send the corrections.

Draws

Install

The marketplace name is chien-mods.

In your shell
claude plugin marketplace add vumichien/claude-code-mods-kit
claude plugin install assumption-check@chien-mods
Open the repo

It runs inside the Claude Code process. The side panel is only what the author's README names. Read the repo before you install, or run claude plugin validate on the plugin folder.

Version
0.1.1
License
MIT
Repo stars
4
Last change
2026-10-10

What it can reach

  • Draws in the session
  • Sees tool calls

Start a new session after installing, then turn it on with /assumption-check on; it starts off. Tested on Claude Code 2.1.296 on Windows 11; mods are an early-access feature, so the API can change between versions. It checks a turn that edited a file or ran a command, not one that only read or one you interrupted, and the check starts after the turn has ended, so the turn never waits for it. It asks Haiku for at most 8 assumptions. The list lives for the session only. It lists what it finds and does not prove the list complete. It has no options. To try it for one session without installing, git clone the repository and start claude with --plugin-dir pointing at the mod's folder.

The marketplace, chien-mods, also holds secret-guard, plan-meter, done-gate, context-meter, idea-shelf, session-monitor, and prompt-enhancer, each installed on its own. Checking Claude's unstated guesses against what you meant is why this page files it under review. decision-lens, by theonly1me, lists the decisions Claude made in each turn in a /why pane, with why and what it chose not to do. quiz-after, by DannyMac180, draws a quiz about the change above the prompt when Claude finishes a task that changed something. This one asks Haiku for a list of the choices you did not state, such as a date format, a default, or a library, which shows as a count with a Review button. You mark each Right or Wrong and add a note, and Send corrections puts one prompt in the box saying what to change and what to keep. What Haiku reads is that turn's messages from your last prompt on, each tool call's input cut to 1,500 characters, the whole at most 30,000, after secret-guard, if on, has hidden what it hides.

Demo

The assumption-check paragraph of the author's claude-code-mods-kit README: after a turn that edited a file or ran a command, Haiku gets that turn's messages and returns at most 8 assumptions as JSON; the check starts after the turn ends; Send corrections puts a prompt in the box ("You assumed: ... That is wrong: "); in the author's check a date parser read 3/4/2026 day first, and marking it wrong with the note "our users are in the US" changed the two date formats and nothing else; cost about $0.0008 per checked turn. The examples are the author's.
Screenshot of the author's GitHub page.

What it hooks into

  • session.startWhen a session starts
  • command.runWhen you run its slash command/assumption-check
  • turn.startWhen Claude starts a turn
  • tool.callAround a tool call, where it can hold one
  • turn.completeWhen a turn ends
  • ui.closeWhen its pane closes
  • ui.renderDraws into the sessionAbovePromptPane

Read from the on() calls in its hook modules. Open its source under Files to check what each one does.

Slash commands 1

  • /assumption-checkSwitch the assumption check on or off, show the last check's cost, or open the review

Files

The author's source, 3 files. They load when you open them.

Facts

Author
vumichien
Category
Review
Marketplace
chien-mods
Repo
vumichien/claude-code-mods-kit
Folder
plugins/assumption-check
Version
0.1.1
License
MIT
Repo stars
4
Last change
2026-10-10
Language
TypeScript
Files
7

Other mods in Review

replay-theaterAnthropicAdds /replay. Steps through the files Claude edited last turn, one diff at a time.review-watchhamzaferDraws one line above the prompt per running code review, with the model, the target, the elapsed time, and Codex's latest output. A toast lists the findings.md-previewhamzaferWhen Claude edits Markdown, draws the before and after beside the session, rendered the way GitHub does.files-touchedSingh-APDraws one line above the prompt for files Claude changed this session, with lines added and removed. /touched pane lists them, and a press inserts @path.change-ledgerarasovicDraws a pane of every file this session touched, with lines added and removed and which agent edited it, beside the git working tree.turn-timelinearasovicDraws the current turn as a timeline: one lane for the main loop, one per subagent, the split between model, tools, and idle, and the three slowest steps.show-mearasovicDraws each mermaid diagram in an answer as a picture in a pane once the turn ends. /show-me asks for an answer in diagrams.image-peekarasovicDraws thumbnails of images you paste, above the prompt while you write, and larger pictures under the message once it is sent.plannotatorbacknotpropOpens Claude's plan in a review UI where you mark it up, sends your annotations back, and shows what changed when the plan comes back.filetreedata-goblinDraws an IDE-style file tree beside the transcript that lights up what Claude reads, writes, and commits, and follows it through the folders.skinshellosverreRedraws tool rows, code, diffs, tables, math, and Mermaid charts as cards, with a usage band, a calm mode, and fifteen themes; /skin swaps them live.image-viewjarrodwattsDraws thumbnails of the images you paste, in the band above the prompt, instead of bare [Image #1] tags.
All 145 in Review