krzemienski/claude-code-builder
MCP server Claude CodeCodexCursor +2
MCP server "context7" as configured in krzemienski/claude-code-builder. Runs locally from the @context/mcp npm package.
krzemienski/claude-code-builder
MCP server Claude CodeCodexCursor +2
MCP server "memory" as configured in krzemienski/claude-code-builder. Runs locally from the @memory/mcp npm package.
krzemienski/claude-code-builder
MCP server Claude CodeCodexCursor +2
MCP server "sequential-thinking" as configured in krzemienski/claude-code-builder. Runs locally from the @sequential-thinking/mcp npm package.
krzemienski/claude-code-builder
MCP server Claude CodeCodexCursor +2
MCP server "filesystem" as configured in krzemienski/claude-code-builder. Runs locally from the @filesystem/mcp npm package.
krzemienski/claude-code-builder
MCP server Claude CodeCodexCursor +2
MCP server "git" as configured in krzemienski/claude-code-builder. Runs locally from the @git/mcp npm package.
krzemienski/claude-code-builder
Instructions file
Instructions for krzemienski/claude-code-builder, covering claude.md, project overview, 🚨 critical: mcp server usage is mandatory, key development commands and install dependencies.
Plugin Claude Code
Plugin marketplace listing 1 plugin: proofpunk.
Instructions file CodexOpenCode
AGENTS.md instructions for krzemienski/proofpunk, covering proofpunk, purpose, key files, subdirectories and for ai agents.
Instructions file
Claude Code instructions for krzemienski/proofpunk: Generated artifacts — never hand-edit; change the source and regenerate.
Plugin Claude Code
Execution-first delivery pipeline: brainstorm, plan, decompose into proof-carrying tasks, implement, end-user test, and audit — end-user testing is the only PASS, no mocks, executed evidence over assertion. 18 skills: functional validation, full functional audit, validation planning with cumulative proof obligations…
Agent
Drives the real system as the end user to prove a completed change — browser clicks, real curl payloads, real CLI invocations, simulator, or PTY for TUIs. Produces run-scoped evidence and a per-criterion verdict. Use after any production-code change and before any 'done' claim.
Agent
End-user-driven builder subagent. Runs the proofpunk write path in its own context — scouts the real codebase, implements production code only, then drives the finished feature as the end user with run-scoped evidence. Never writes test files. Use for any 'implement/build/ship X' delegation.
Agent
Read-only codebase explorer. Walks the real tree before any edit — structure, patterns, contracts, touchpoints — and returns a 3-6 bullet context summary with named files. Use at the start of any implement run and before any non-trivial change.
Command
Author a new high-quality prompt on the canonical XML skeleton.
Command
Implement a goal end to end with the Proofpunk orchestrator.
Command
Install proofpunk project memory — CLAUDE.md or AGENTS.md (platform-correct, ≤200 lines, merged never clobbered) plus scoped rules.
Command
Rate a prompt file against the 100-point rubric and apply remediations to file.
Command
Run a repo-wide intent-vs-code truth audit with evidence-backed findings.
Command
End-user test the current work as the end user, producing executed proof.
Hook
Runs when a session starts on startup, resume and clear, executing session-start.sh via sh. From krzemienski/proofpunk.
Hook
Runs when the agent finishes a response, executing stop-guard.sh via sh. From krzemienski/proofpunk.
Hook
Runs when a subagent finishes, executing stop-guard.sh via sh. From krzemienski/proofpunk.
Hook
Runs before the agent uses a tool for Write and Edit tool calls, executing no-test-files.sh, evidence-guard.sh and capture-guard.sh via sh (3 commands). From krzemienski/proofpunk.
Hook
Runs after instructions files are loaded, executing instructions-loaded.sh via sh. From krzemienski/proofpunk.