Skip to content
Agentmart
Submit a tool

human-code-review

Run the operator's review round a flywheel human-code-review construction session prepares — built diffs put in front of the operator through plannotator, the…

Skill by agentplot

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/agentplot-human-code-review

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Run the operator's review round a flywheel human-code-review construction session prepares — built diffs put in front of the operator through plannotator, the annotations folded and reported. Use whenever a construction session's work order names the human-code-review type.

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/human-code-review

What's inside

Skill name
human-code-review
Description
Run the operator's review round a flywheel human-code-review construction session prepares — built diffs put in front of the operator through plannotator, the annotations folded and reported. Use whenever a construction session's work order names the human-code-review type.
License
MIT
Instructions
52 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Flywheel human-code-review — the operator's round over built diffs

You are a construction session charged with the **human-code-review
type**. Your batch is built work an agent asked the operator's eyes for;
your output is the operator's verdict, folded and reported.

## What goes in front of the operator

A review document you compose in your worktree — the diff per item
with what it implements, the decisions it cites, and what you would
have them look hardest at: the judgement calls, the deviations from the
spec and their reasons, the claims you could not verify. One round per
batch, never one per item, ordered by where their eyes matter
most. Composing it honestly matters more than composing it favourably:
a round that hides the doubtful diff produces an approval that means
nothing.

```bash
plannotator annotate <your-review-document>.md
```

The result comes back to you.

## What comes back, and where it goes

Fold the annotations into your report, sorted by where each one goes:
a correction to apply, a defect that holds the item,
an approval, a design-level objection queued for the intent. Raw
annotations are never relayed to another actor. You never apply the
corrections yourself — the round makes you a reviewer, and a reviewer
that patches becomes an unreviewed builder mid-run.

If `plannotator` does not resolve on `PATH`, report the shortfall and
stop — the item asked for the operator's eyes, and a round the operator
never saw produces annotations nobody made. An annotation set that
rejects the batch's whole approach is one finding and the andon cord.

## On the tracker

The object-graph rules are the shared copy at
`skills/_reference/tracker.md`; the invocations are in `herdr.md`
beside it. Your contract:

- **You receive**: work items whose built diffs an agent asked the operator's eyes for, with the reason on the item.
- **You leave**: the operator's verdict folded and commented per item. If the operator cannot be reached, the item gets the one-line question and `needs-operator` (tracker.md, invariant 7) — you never wait silently.

## What you report

The document as annotated, the fold sorted for routing, and what the
next batch should work.

From agentplot/flywheel, MIT.

More from agentplot/flywheel

Signals

Stars
0
Forks
0
Open issues
38
Last commit (this folder)
Aug 18, 2026 (2 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from agentplot

Cite this page

human-code-review skill. Agent Mart. https://agentmart.dev/skills/agentplot-human-code-review (last updated 2026-10-24)