By Praneeth-496
Portable Claude Code context, 30 skills, 14 agents (review + verification/anti-hallucination + ideation), safety hooks, a /toolkit-init bootstrap command, and a bundled zero-dependency MCP server (verify_result_claim, query_graph, memory_graph_add, placeholder_scan). File-based + graph memory, council/debate orchestration, Chain-of-Verification + citation auditing, divergent->convergent brainstorming, HPC/SLURM helpers, devcontainer-ready security.
Admin access level
Server config contains admin-level keywords
Executes bash commands
Hook triggers when Bash tool is used
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Steelmans the opposite of the proposed change. Use before irreversible decisions or merging anything labelled "obviously correct". Read-only.
Surfaces the IMPLICIT assumptions a claim, plan, or design silently depends on, using the Toulmin argument model. Finds the unstated "this only works if ___" conditions that, if false, sink the conclusion. Complements flaw-level critique (which attacks what is written) by exposing what is NOT written. Use before betting on a plan or accepting an argument. Read-only. Project-agnostic.
Per-claim source-grounding auditor. For every factual sentence in a document, answer, or code comment, retrieves supporting evidence and labels it Supported / Unsupported / Contradicted with a citation. Catches invented papers, wrong DOIs, fabricated APIs, and plausible-but-unsourced numbers. Use on any text that asserts external facts or cites sources. Read-only. Project-agnostic.
Strict code reviewer. Use after a meaningful diff exists (>5 lines or >1 file). Reviews staged or recently-edited code for correctness, security (OWASP), readability, and over-engineering. Read-only.
Chain-of-Verification fact checker. Takes a claim, answer, or generated passage and reduces hallucination by generating independent verification questions, answering each in ISOLATION (so the original wording can't bias the check), then flagging or revising anything the checks contradict. Use on any high-stakes factual output before it is acted on. Read-only. Project-agnostic.
Rewrite text to be dramatically sharper — tighter, more vivid, more memorable. TRIGGER when user says "make this sharper", "tighten this", "rewrite better", "make it punchier", "improve this paragraph", "this is too wordy", or pastes prose/commit-msg/PR-text and asks for a better version. For thesis paragraphs, abstracts, copy, commit messages.
Bootstrap or refresh the project memory directory automatically by scanning the current repo. Derives a project snapshot, a code-style snapshot, and a reference-paths snapshot from what's actually on disk — no user input required. Run on a fresh project to seed memory; re-run any time to rebuild from current state.
End-to-end divergent->convergent ideation pipeline. Reframes the problem, generates diverse options, stress-tests them, and converges to a recommended shortlist — wiring the ideator, premortem, assumption-surfacer, and synthesizer agents into one disciplined pass. Use for "help me brainstorm", "explore options for X", "what are our approaches and which is best". Project-agnostic.
One-sentence-or-paragraph answer, zero preamble. TRIGGER when user says "in one line", "one sentence", "TL;DR", "short answer", "quick", "just the answer", "don't explain", or the question is a factual lookup that doesn't need elaboration. Style change — apply for one turn.
Side-by-side comparison table + verdict. TRIGGER when user says "X vs Y", "compare A and B", "which should I use", "should I pick X or Y", "what's better, A or B", or names two-or-more concrete alternatives (libraries, models, approaches, file structures, designs) and asks to choose.
Modifies files
Hook triggers on file write and edit operations
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
▶ Open the live dashboard · View full-size screenshot
Portable Claude Code context, skills, and guard rails. Drop the claude-toolkit/ folder onto any machine, run install.sh in a project, and that project gets the same disciplined setup: a tiny always-loaded CLAUDE.md, a lazy-loaded CONTEXT.md, sane permissions, and a curated set of user-scope skills.
Current version: see claude-toolkit/VERSION.
# 1. Clone the toolkit somewhere stable (one-time)
git clone <this-repo> ~/Documents/claude-toolkit
# 2. Bootstrap a project
cd /path/to/your/project
bash ~/Documents/claude-toolkit/claude-toolkit/install.sh
# Note: auto mode is enabled by default (requires Team/Enterprise/API plan).
# Disable: remove "defaultMode":"auto" from .claude/settings.local.json
# 3. Fill in the placeholders the installer flagged
grep -nE '<[A-Z_]+>' .claude/CLAUDE.md .claude/CONTEXT.md
# 4. (Optional) Let Claude bootstrap project memory by scanning the repo
# Open Claude Code in the project and say: "run auto-memory"
That's it. Claude Code will pick up .claude/CLAUDE.md on the next session and the user-scope skills are now available in every project.
bash ~/Documents/claude-toolkit/claude-toolkit/install.sh # safe: skips files that already exist
bash ~/Documents/claude-toolkit/claude-toolkit/install.sh --update # overwrite generic skills with newer toolkit copies (templates left alone)
bash ~/Documents/claude-toolkit/claude-toolkit/install.sh --force # also overwrite project templates (DESTRUCTIVE — backs nothing up)
The default mode is idempotent: safe to re-run any time.
claude-toolkit/
├── README.md this file
├── claude-toolkit/
│ ├── VERSION toolkit version (stamped into ~/.claude/skills/.claude-toolkit-version on install)
│ ├── install.sh bootstrap script (idempotent; --update / --force)
│ ├── .claude-plugin/
│ │ ├── plugin.json Claude Code plugin manifest (for /plugin install)
│ │ └── marketplace.json marketplace entry pointing back at this repo
│ ├── agents/ project-scope subagents installed to .claude/agents/
│ │ ├── code-reviewer.md strict diff review (sonnet, read-only)
│ │ ├── security-auditor.md OWASP-style security pass on diff (sonnet, read-only)
│ │ ├── test-runner.md detects + runs project's test command (sonnet)
│ │ ├── simplifier.md flags dead code, single-use abstractions (sonnet, read-only)
│ │ ├── adversary.md steelman the opposite of the proposed change (opus)
│ │ ├── doc-writer.md updates docs to match a code change (sonnet, docs-only)
│ │ ├── fact-checker.md verifies individual claims (paths, symbols, numbers, signatures) against the repo (sonnet, read-only)
│ │ └── flow-auditor.md verifies the *reasoning chain* between claims (catches unsupported leaps) (opus, read-only)
│ ├── templates/
│ │ ├── CLAUDE.md.template always-loaded project onboarding (≤150 lines target)
│ │ ├── CONTEXT.md.template lazy-loaded heavier background
│ │ ├── settings.local.json.template permissions + deny list + hooks + statusline
│ │ ├── statusline.sh toolkit statusline: branch | dirty | model | cost (ccusage)
│ │ ├── gitignore-snippet.txt lines appended to a project .gitignore
│ │ ├── hooks/ deterministic guardrails (29 events at code.claude.com/docs/en/hooks)
│ │ │ ├── block-dangerous-bash.sh PreToolUse(Bash): block force-push, rm -rf /, fork bombs, etc.
│ │ │ ├── format-on-write.sh PostToolUse(Edit|Write): run ruff/prettier/rustfmt etc.
│ │ │ ├── session-briefing.sh SessionStart: inject git briefing
│ │ │ ├── pre-compact-briefing.sh PreCompact: snapshot session state
│ │ │ └── post-compact-restore.sh PostCompact: re-inject snapshot
│ │ ├── devcontainer/ sandboxed VS Code devcontainer for bypass-mode work
│ │ │ ├── devcontainer.json Ubuntu 24.04 + Claude Code, no Docker socket / SSH mounts
│ │ │ └── post-create.sh installs ruff/prettier/shfmt/ccusage
│ │ └── github-workflows/ drop into .github/workflows/
npx claudepluginhub praneeth-496/claude-toolkit --plugin claude-toolkitHarness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns