research-tools

Skills

AI agent skills collection — a Claude Code plugin marketplace.

Install

/plugin marketplace add soilSpoon/skills
/plugin install slice@soilspoon-skills           # trust-first engineering workflow (+5 role agents)
/plugin install dev-toolkit@soilspoon-skills      # frontend fundamentals, build-config drift, root-cause workflow
/plugin install job-hunt@soilspoon-skills         # resume tailoring, Wishket applications
/plugin install research-tools@soilspoon-skills   # fact-checking, browser automation

Individual skills also install standalone (any SKILL.md-compatible tool — Claude Code, opencode, …):

npx skills add soilSpoon/skills@<skill-name>

Contributing or growing a skill? See docs/skill-hygiene.md for size budgets and structure rules that keep this repo lean.

Available Skills

tailor-resume

Tailor resume and career description for a specific job posting.

npx skills add soilSpoon/skills@tailor-resume

What it does:

• Analyzes job posting requirements
• Filters relevant experience/skills from master resume data
• Generates a tailored variant
• Evaluates via Groupby API (서류 합격률 예측 + 이력서 강화 팩폭)

Bundled scripts:

• scripts/groupby-api.mjs — Groupby resume analysis API client (no browser needed)

slice

Trust-first recursive decomposition of a development task — Kent Beck's Mastering Programming heuristics (Slicing, Baseline Measurement, One-thing-at-a-time) as an executable workflow.

npx skills add soilSpoon/skills@slice

What it does:

• Pins a deterministic baseline (tests + invariants at a git SHA) before any work begins
• Recursively classifies the task (easy/hard × small/big) into thin vertical slices
• Executes each atomic leaf with Canon TDD (call-your-shot, one test at a time, two hats)
• Adversarially verifies every leaf (executor ≠ verifier, deterministic gates before model judgment, risk-tiered lenses)
• Self-repairs while issues strictly converge; halts the approach on an untrusted streak
• Lands one git commit per trusted leaf; deterministic full-suite gate at integrate
• Ends with an Owner's Briefing — a guided read to repay comprehension debt
• Optional parallel mode: isolated git worktrees per independent slice group (+ shared build scratch for compile-bound repos)

Bundled:

• recursive-slice.js — the workflow engine (run via the Workflow tool's scriptPath)
• agents/slice-*.md — the 5 roles (baseliner/assessor/slicer/executor/verifier) as standalone subagents
• references/ — architecture, philosophy, and 11 battle-tested lessons (deadlocks, orphaned test runners, false greens)
• scripts/ — live run viewer, no-focus window capture for visual UI verification (macOS)

fact-check

Fact-check articles, newsletters, and written content by dispatching 6 parallel verification agents.

npx skills add soilSpoon/skills@fact-check

What it does:

• Accepts text, file path, or URL as input
• Extracts claims and classifies risk (high/medium/low)
• Dispatches 6 parallel agents: source verify, number check, freshness check, context check, link check, assumption check
• Cross-references results with Tier 1-4 source credibility system
• Auto-searches for higher-tier sources when only Tier 3-4 found
• Outputs structured report with verdicts, conflicts, and diff-style fix suggestions

human-like-browser

Human-like browser automation that bypasses bot detection systems.

npx skills add soilSpoon/skills@human-like-browser

What it does:

• Bezier-curve mouse movements with micro-tremor and Fitts's Law timing
• Log-normal typing delays with digraph timing and typo simulation (~600 chars/min)
• Smooth inertia scrolling with trackpad/mousewheel mix
• FCaptcha keystroke biometrics bypass (autocorrelation, dwell variance, log-normal fit)
• Comprehensive anti-fingerprinting stealth (navigator.webdriver, hardware, WebGL, CDP artifacts)
• Auto-stealth on navigation, idle fidget, element-scoped scrolling

Bundled scripts:

• scripts/human-behavior.js — Initialization block reference (paste into Playwright browser_run_code)

apply-wishket

Generate tailored proposals for Wishket freelance projects.

npx skills add soilSpoon/skills@apply-wishket

What it does:

• Analyzes project requirements from Wishket URL or project ID
• Matches relevant experience from master.yaml
• Estimates cost and timeline with multi-step validation
• Generates proposal with issue→solution pattern and staged process
• Fact-checks output via parallel verification agents
• Supports single or batch (parallel) project processing

Bundled scripts:

• scripts/verify-proposal.sh — 12-point structural validation for generated proposals

toss-frontend-fundamentals

Toss의 Frontend Fundamentals 코드 품질 4대 기준(가독성/예측가능성/응집도/결합도) + 접근성(A11y)을 반영하는 프론트엔드 코드 리뷰 & 작성 가이드.