Coding & Development✨gemini🤖claude🌐universal

Shared Dotfiles & Configs

Personal collection of generalizable dotfiles, configurations, and Claude Code skills for reuse across projects.

Quick Start

Install this Skill

npx skills add isnbh0/shared
# Shared Dotfiles & Configs

Personal collection of generalizable dotfiles, configurations, and Claude Code skills for reuse across projects.

[한국어](README.ko.md)

## Installation

### Option A: Plugin (recommended)

First, register the marketplace:

```bash
/plugin marketplace add isnbh0/shared
```

Then install individual skills:

```bash
/plugin install interview@isnbh0
/plugin install spex@isnbh0
/plugin install critique@isnbh0
/plugin install macros@isnbh0
/plugin install study@isnbh0
/plugin install gimme@isnbh0
```

### Option B: Symlink / Copy

```bash
# Clone to your preferred location
git clone <repo-url> ~/shared

# Symlink .claude directory into a project
ln -s ~/shared/.claude /path/to/project/.claude

# Or copy specific skills
cp -r ~/shared/plugins/spex/skills/spex /path/to/project/.claude/skills/
```

## Skills

### Published (marketplace)

Installable via `/plugin install <name>@isnbh0`:

#### critique

External AI critique via CLI tools (Codex, Gemini).

```
/critique:codex [file-path] [focus]
/critique:gemini [file-path] [focus]
```

- Gets an independent second opinion on specs, code, or recent changes
- **Codex backend:** Runs in a read-only sandbox, supports reasoning effort levels (default, high, xhigh)
- **Gemini backend:** Sandboxed with write and network restrictions

#### interview

Structured requirements discovery through conversational interviews.

```
/interview <topic> [--ref <path>] [--workspace <dir>]
```

- Extracts requirements, constraints, and design decisions through guided Q&A
- Supports reference files to anchor discussion around existing artifacts
- Produces timestamped synthesis documents in the workspace

#### study

Document-grounded Socratic study sessions on any URL or local file.

```
/study <uri>
```

- Calibrates to user familiarity (full study, guided study, or gap check)
- Persists session notes to markdown files with resumption across sittings
- Configurable sessions directory and optional freeform `instructions` for persona/behavior customization

#### spex

Two-phase specification and implementation workflow that separates planning from execution.

Three self-contained skills:

- `/spex:write` — Create a spec, commit it, and stop — no implementation
- `/spex:write-phased` — Create a multi-phase spec for complex features
- `/spex:implement` — Follow an existing spec, implement, update status, and commit

#### macros

Subagent orchestration workflows and session modes: map-reduce, research-backed critique, consensus review, sequential passes, and rigor mode.

```
/macros:mapreduce <task> [--workspace <dir>]
/macros:doubt ["freeform question"]
/macros:consensus <count>
/macros:seq <count>
/macros:rigor
/macros:askme
/macros:delegate
/macros:tmi
/macros:dry-run
/macros:timeless
/macros:dredge ["freeform query"]
```

- **mapreduce** — Splits tasks into independent chunks, dispatches parallel subagents, consolidates results
- **doubt** — Spawns a blind subagent that reads code, verifies assumptions against web sources, applies fixes, and reports concerns ranked by severity
- **consensus** — Runs N blind agents on the same job in parallel, merges findings into consensus/unique/conflicts (no edits for concurrent safety)
- **seq** — Runs N serial blind passes with commits between rounds; requires clean worktree
- **rigor** — Activates Rigor Mode for the session: prioritizes correctness, thorough investigation, and web-grounded verification over minimalism
- **askme** — Shorthand: stop and ask the user to clarify ambiguities or make decisions instead of assuming
- **delegate** — Shorthand: prefer subagents to save context space and parallelize independent subtasks
- **tmi** — Flags content that only makes sense if you were in the room when it was written; reports by default, edits if explicitly instructed
- **dry-run** — One-shot failsafe: describes what it would do for the next request instead of doing it, then waits for confirmation
- **timeless** — Shorthand: avoid time estimates (hours, calendar, size-to-time buckets); describe complexity, scope, risk, and ordering instead
- **dredge** — Searches prior Claude Code chat transcripts under `~/.claude/projects` for context; defaults to the current project, widens scope and time window from natural-language hints in the query (e.g. "across all projects", "in the craken repo", "yesterday")

#### gimme

User-invoked inversion of delegation — hand Claude a request and get back a filesystem bundle you can act on.

```
/gimme
```

- Writes a timestamped bundle with `checklist.md`, `notes.md` (template with pre-labeled paste slots), and an empty `dropbox/` directory for file artifacts
- Each checklist item has action / why-it's-on-you / drop-path so results land somewhere Claude can pick up without further direction
- Optional `launch_command` config (e.g. `cursor {path}`, `code {path}`, `open {path}`) opens the bundle in your editor immediately
- Never self-invoked — only runs when you explicitly call `/gimme`

### Oth

... (truncated)
Repository
3stars
1forks
View on GitHub
Details
DiscoveredApr 22, 2026
Last UpdatedApr 22, 2026
Author@isnbh0
Quality Score
35/100