/human
Request Ben's own review of the working diff or a document before it leaves the machine, for architecture and slop rather than bugs. Runs after the mechanical passes (review:code or simplify, comments:audit). Use for "get my review", "let me look at this first", or as the last
$ npx -y skills add bendrucker/claude --skill human --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
- Slash command
/human
Context preview
The summary Claude sees to decide when to auto-load this skill.
Request Ben's own review of the working diff or a document before it leaves the machine, for architecture and slop rather than bugs. Runs after the mechanical passes (review:code or simplify, comments:audit). Use for "get my review", "let me look at this first", or as the last
SKILL.md
human.SKILL.mdname: review:human
description: >-
Request Ben's own review of the working diff or a document before it leaves the machine, for architecture and slop rather than bugs. Runs after the mechanical passes (review:code or simplify, comments:audit). Use for "get my review", "let me look at this first", or as the last pre-PR pass.
argument-hint: "[--doc <file>] [--browser] [--summary <text>]"
allowed-tools:
- "Bash(bun ${CLAUDE_PLUGIN_ROOT}/scripts/attention.ts:*)"
- Bash(plannotator review:*)
- Bash(plannotator annotate:*)Human Review
Put the change in front of Ben and act on what comes back. `attention.ts` labels the herdr pane `review`, sets its `$review` token, and raises a toast. Outside herdr, `raise` and `clear` are no-ops and `open` exits 1.
Context
- herdr pane: !`[ "$HERDR_ENV" = 1 ] && printf %s "${HERDR_PANE_ID:-none}" || printf none`
Modes
| Mode | Surface | Turn | |---|---|---| | Terminal (default), diff | reviewr sidebar over the working diff | Ends. Comments arrive as the next user turn | | Terminal (default), `--doc <file>` | plannotator-tui over the file | Ends. Annotations arrive as the next user turn | | `--browser`, diff | `plannotator review --git --json [--base <ref>]` | Blocks. The decision is the tool result | | `--browser`, `--doc <file>` | `plannotator annotate <file> --gate --json --require-approval` | Blocks. The decision is the tool result |
Terminal surfaces open as a split beside this pane, named explicitly and without taking focus, and deliver into it once it is idle. So the turn has to end before Send can land. They need the herdr pane above. When it reads `none`, run the gate as if `--browser` were passed.
Precondition
Request after the mechanical passes have run. `/ship` orders them. Standalone on an unreviewed diff, say so in the `--summary` text rather than running them here.
Request
1. `bun ${CLAUDE_PLUGIN_ROOT}/scripts/attention.ts raise --summary "<repo> <branch>: <what to review>"`. A caller's `--summary` passes through verbatim. 2. `bun ${CLAUDE_PLUGIN_ROOT}/scripts/attention.ts open --diff`, or `open --doc <file>` for a document. `open --diff` re-uses a reviewr sidebar already up over this working tree. 3. End the turn with exactly one line naming the surface and that Send resumes the work. The surface shows the change. The line carries no recap.
Resume
The next user turn carries the comment batch plus Ben's own remarks. Each block is a `path:lines` header, the verbatim snippet, and his text. The herdr plugin's `skills/herdr/references/reviewr.md` covers the shape in full. Act on all of it. Sending cleared reviewr's store. The batch in the turn is the only copy. This plugin's `UserPromptSubmit` hook has already cleared the pane label.
The review ends when Ben says approve, ship, or no further comments. Otherwise act on the batch and request again.
Gate
With `--browser`:
1. `attention.ts raise --summary "..."` 2. `plannotator review --git --json`, adding `--base <ref>` when the caller resolved one, or `plannotator annotate <file> --gate --json --require-approval` 3. `attention.ts clear` 4. Act on the decision. `approved` ends the review. `annotated` means act on the annotations and run the gate again. `dismissed`, or exit 1 from the annotate gate, means not approved: fold the feedback in and run the gate again. Exit 2 is a startup error: report it and stop.
Read more
name: review:human
description: >-
Request Ben's own review of the working diff or a document before it leaves the machine, for architecture and slop rather than bugs. Runs after the mechanical passes (review:code or simplify, comments:audit). Use for "get my review", "let me look at this first", or as the last pre-PR pass.
argument-hint: "[--doc <file>] [--browser] [--summary <text>]"
allowed-tools:
- "Bash(bun ${CLAUDE_PLUGIN_ROOT}/scripts/attention.ts:*)"
- Bash(plannotator review:*)
- Bash(plannotator annotate:*)Human Review
Put the change in front of Ben and act on what comes back. `attention.ts` labels the herdr pane `review`, sets its `$review` token, and raises a toast. Outside herdr, `raise` and `clear` are no-ops and `open` exits 1.
Context
- herdr pane: !`[ "$HERDR_ENV" = 1 ] && printf %s "${HERDR_PANE_ID:-none}" || printf none`
Modes
| Mode | Surface | Turn | |---|---|---| | Terminal (default), diff | reviewr sidebar over the working diff | Ends. Comments arrive as the next user turn | | Terminal (default), `--doc <file>` | plannotator-tui over the file | Ends. Annotations arrive as the next user turn | | `--browser`, diff | `plannotator review --git --json [--base <ref>]` | Blocks. The decision is the tool result | | `--browser`, `--doc <file>` | `plannotator annotate <file> --gate --json --require-approval` | Blocks. The decision is the tool result |
Terminal surfaces open as a split beside this pane, named explicitly and without taking focus, and deliver into it once it is idle. So the turn has to end before Send can land. They need the herdr pane above. When it reads `none`, run the gate as if `--browser` were passed.
Precondition
Request after the mechanical passes have run. `/ship` orders them. Standalone on an unreviewed diff, say so in the `--summary` text rather than running them here.
Request
1. `bun ${CLAUDE_PLUGIN_ROOT}/scripts/attention.ts raise --summary "<repo> <branch>: <what to review>"`. A caller's `--summary` passes through verbatim. 2. `bun ${CLAUDE_PLUGIN_ROOT}/scripts/attention.ts open --diff`, or `open --doc <file>` for a document. `open --diff` re-uses a reviewr sidebar already up over this working tree. 3. End the turn with exactly one line naming the surface and that Send resumes the work. The surface shows the change. The line carries no recap.
Resume
The next user turn carries the comment batch plus Ben's own remarks. Each block is a `path:lines` header, the verbatim snippet, and his text. The herdr plugin's `skills/herdr/references/reviewr.md` covers the shape in full. Act on all of it. Sending cleared reviewr's store. The batch in the turn is the only copy. This plugin's `UserPromptSubmit` hook has already cleared the pane label.
The review ends when Ben says approve, ship, or no further comments. Otherwise act on the batch and request again.
Gate
With `--browser`:
1. `attention.ts raise --summary "..."` 2. `plannotator review --git --json`, adding `--base <ref>` when the caller resolved one, or `plannotator annotate <file> --gate --json --require-approval` 3. `attention.ts clear` 4. Act on the decision. `approved` ends the review. `annotated` means act on the annotations and run the gate again. `dismissed`, or exit 1 from the annotate gate, means not approved: fold the feedback in and run the gate again. Exit 2 is a startup error: report it and stop.
My personal plugin marketplace for Claude Code, Anthropic's AI coding assistant.
Repo: bendrucker/claude
Other skills on bendrucker-claude.
cleye
Type-safe CLI argument parsing with cleye, the standard parser for this repo's Bun scripts. Use when writing or editing any script that takes arguments (flags,…
activity
Report real device usage from ActivityWatch. Covers per-app time, window titles, and active vs idle spans. Use when asked "what apps did I use", "how long was…
history
Report shell history from atuin's local capture. Covers what commands ran, when, where, and how they exited. Use when asked "what commands did I run", "what…
bun
Bun runtime patterns. Use when running bun commands, working with package.json/bun.lock, writing TypeScript scripts under Bun, or developing Claude Code…
agent-team
Orchestrating Claude Code agent teams. Use when creating teams, spawning teammates, assigning tasks, configuring teammate modes, or setting up team quality…

