DOCSEARCHbyAlgolia
Submit Docs

Build with Claude

repository·main·Indexed 25 days ago

https://github.com/davepoon/buildwithclaude

A plugin marketplace and discovery platform for Claude Code that provides curated collections of Agents, Commands, Hooks, and Skills. It includes specialized AI subagents such as Agent Triforce (Prometeo, Forja, and Centinela) for PM/Dev/QA workflows, uc-taskmanager for multi-agent pipeline orchestration, AIBoarding for AI agent context management via AIBOARDING.md, and aigent-os for session-lifecycle management using a Markdown vault.

Tokens
318.6K
Snippets
619
Records
1.3K
Agent score
83%

What's inside buildwithclaude

  1. Overview of Ultraship capabilities

    main

    Ultraship provides three categories of capabilities for building and scaling production software:

    Builder Skills (32)

    Includes TDD enforcement, code review with confidence scoring, structured planning, frontend/API design, data modeling, deploy pipelines with rollback plans, debugging, refactoring, git workflow, competitive analysis, incident response, and launch prep.

    Audit Tools (29)

    Includes secret scanners, code profilers (N+1 queries, memory leaks, ReDoS), bundle trackers, dependency doctors (import graph analysis), Lighthouse runners, health checks (HTTP + SSL + security headers), env validators, migration checkers, and SEO/GEO/AEO scanners (60+ rules).

    Agents (9)

    Includes parallel dispatching for /ship scorecard generation, code review, security audit, performance audit, browser verification, competitive analysis, launch prep, incident response, and growth tracking.

  2. Overview of erne-universal components

    main

    The erne-universal plugin includes the following specialized components for mobile development:

    • 8 Agents: Specialized AI experts for RN/Expo development.
    • 16 Commands: Slash commands for common mobile development tasks.
    • 8 Skills: Reusable capabilities for mobile workflows.
    • 10 MCP Integrations: Tool and service connections.

    Capabilities are categorized under mobile-development, react-native, and ai-agents.

  3. Overview of Agent Triforce Agents

    main

    Agent Triforce uses three specialized agents to coordinate development workflows:

    • Prometeo (PM): Focuses on product management, including feature specs, user stories, business logic, prioritization, and roadmaps.
    • Forja (Dev): Focuses on technical execution, including architecture, implementation, testing, infrastructure, and documentation.
    • Centinela (QA): Focuses on quality and security, including code review, security audits, compliance, dead code detection, and release gates.

  4. Overview of ThumbGate features

    main

    ThumbGate provides safety and compliance layers for AI coding agents through the following features:

    • Pre-action gates: 33 specific gates to intercept destructive operations.
    • Feedback loop: Thumbs up/down mechanism for continuous improvement.
    • Budget enforcement: Controls to manage usage costs.
    • Compliance: Includes NIST/SOC2 compliance tags.

  5. Overview of ultracost features

    main

    The ultracost plugin provides per-stage model routing for Claude Code ultracode dynamic workflows. It ensures cost-effective execution by applying a quality-first routing policy to subagents.

    Key features include:

    • Quality-first policy: Routes coding/reasoning to opus @ xhigh and mechanical/search work to sonnet. haiku is never used.
    • Always-on guidance: A SessionStart hook injects routing policies into the session.
    • The Workflow Guard: A static analyzer that flags agent() stages missing a model: pin, using banned models, or mismatching work descriptions (errors UC001–UC008).
    • Pre-flight cost gate: A PreToolUse hook that estimates workflow costs and requires user approval (Approve / Modify / Cancel) before execution.
    • Closed loop: Reconciles estimated vs. actual costs using local transcripts to self-calibrate the estimator and maintain a savings ledger.

  6. Claude HUD features overview

    main

    Claude HUD provides a real-time statusline HUD for Claude Code with the following capabilities:

    • Session Info: Displays model name, color-coded context usage bar, config counts, and session duration.
    • Tool Activity: Shows running tools with a spinner and counts of completed tools.
    • Agent Tracking: Tracks active subagents including their descriptions and runtime.
    • Todo Progress: Displays the current task and completion metrics.

  7. GSD Plugin Features and Capabilities

    main

    The GSD Plugin for Claude Code includes several optimized features:

    • 83 Slash Commands: For planning, execution, debugging, and verification.
    • 33 Agent Definitions: Specialized roles like planner, executor, researcher, verifier, debugger, and UI auditor.
    • MCP Server: Exposes project state as queryable resources and mutation tools.
    • Bundled SDK: Ships its own gsd-sdk binary (v2.42.0+).
    • Auto-resume: Uses a PreCompact hook to write .planning/HANDOFF.json and a SessionStart hook to automatically run /gsd:resume-work after a /compact command.
    • Memory Integration: Persists phase outcomes across sessions via Claude Code's memdir.

  9. Use the ops-doctor skill for health checks and auto-repair

    main

    The ops-doctor skill performs health checks and auto-repairs for the ops plugin. It diagnoses manifest errors, broken permissions, invalid configurations, stale caches, and missing files. It can either report issues via a diagnostic script or spawn an agent to fix them automatically.

    Arguments:

    • --check-only: Run diagnostics and display results without spawning a fix agent.
    • --verbose: Enable detailed output.

  10. Use the claude-code-audit-stack for adversarial verification

    main

    The claude-code-audit-stack provides three specialized subagents designed to catch silent failures in Claude Code workflows:

    1. bot-deploy-verifier: An adversarial post-deploy verifier. It detects the 'silent-skip' pattern (e.g., editing a config file without restarting the service) and prevents accidental cascade restarts of sibling services.
    2. claim-auditor: A quantitative report auditor. It scans Markdown reports for mathematical errors in probability, Expected Value (EV), or Monte Carlo (MC) claims. It flags errors at P1, P2, or P3 severity, specifically looking for:
      • Probability stacking errors (1 − (1 − p)^N vs N × p)
      • Confusion between conditional and marginal pass rates
      • Mixing up percentages and percentage-points
      • Bootstrap-with-replacement implications
      • Best-of-N selection bias
    3. remote-agent-dispatcher: Handles scp-and-spawn for autonomous Claude Code agents on remote hosts. It correctly captures the actual claude binary PID (using pgrep after a fixed wait) rather than the bash wrapper PID.

    Additionally, the stack includes a PostToolUse hook named audit-on-report-write which automatically triggers the claim-auditor whenever a *.report.md file is written.

  11. AIBoarding Lifecycle and Sync Behavior

    main

    AIBoarding manages AI agent context through a lifecycle of creation, synchronization, and updates. It uses deterministic hooks to ensure agents stay aligned with the project's engineering basics and domain logic.

    Lifecycle Stages

    • Create (create-aiboarding): Generates AIBOARDING.md and installs hooks.
    • Sync (session-start / pre-task hooks): Injects the documentation into every session and spawned sub-agents.
    • Update (update-aiboarding): Triages commit diffs to patch drifted sections or advance the pointer.

    Hook Behaviors

    EventHookBehavior
    SessionStart (startup|clear|compact)session-startEmits AIBOARDING.md as session context
    PreToolUse (Task)pre-taskPrepends doc to spawned sub-agents
    PostToolUse (Bash)post-commitCompares last_synced_commit to HEAD, nudges update on drift

  12. Use the ops-triage skill for cross-platform issue triage

    main

    The ops-triage skill performs cross-platform issue triage by pulling data from Sentry (via MCP), Linear (via MCP), and GitHub Issues (via gh CLI). It cross-references issues against the codebase to identify already-fixed issues, auto-resolves them if confirmed, and dispatches agents to handle active issues.

    Arguments:

    • [project-alias|sentry|linear|github|all]: Filters the triage process to a specific project or source.

    Key Capabilities:

    • Gathering: Parallel collection of issues from GitHub, Sentry, and Linear.
    • Cross-referencing: Checks code (via grep, git log, and LSP) and deployment status (e.g., ECS timestamps) to see if an issue is already resolved.
    • Resolution: Automatically closes GitHub issues, updates Linear states to Done, or resolves Sentry issues.
    • Dispatching: Spawns specialized fix agents to address active issues.