ccgm

Adversarial review of plans and artifacts with personal lead review by default and bounded opposite-provider review on explicit request. A designated writer applies authorized fixes; report-only targets stay unchanged.

Session posture for expensive orchestrator models (Fable/Opus): /advisor puts the main agent into a delegation-only mode - it writes specs, dispatches cheaper implementer agents, personally reviews spec compliance and quality by default (cross-provider review is opt-in), triages findings, and merges, while a hard PreToolUse gate (exit 2, bypass-surviving) blocks its own file edits and non-orchestration Bash. Subagent calls pass untouched; a per-turn posture injection keeps the model delegating instead of fighting denials. ADVISOR_DIRECT=1 one-off hatch. State is per session: every session starts in the mode, and CCGM_ADVISOR_AUTO=false opts out.

[DEPRECATED] Go-based terminal UI for managing Claude Code agent processes via tmux panes. Unmaintained and no longer offered for new installs; a GUI-based replacement is being pursued instead. Kept in-repo for existing users.

Principles and audit skill for building applications where an agent is a first-class user. Four principles (parity, granularity, composability, emergent capability) as a rule file, a self-eval / red-team rubric that scores a surface against those principles, plus /agent-native-audit that scores a codebase with specific counts and a reviewer persona for the unified review orchestrator.

/argus - a closed-loop visual-ATDD harness where the agent develops UI against a per-feature design spec and signs off on its own work (functional AND visual) by grounding every judgment in deterministic gates plus a SEPARATE judge subagent that scores the render against the spec, a reference image, and the design system — never the diff. Loops to two consecutive passes, then commits a snapshot baseline. Platform-agnostic via a pluggable sensor+gates adapter (web built in; iOS/macOS via a project adapter). Ships the loop skill, the argus-judge agent, the spec/verdict/gate/rubric schemas, six dependency-free deterministic gate scripts, and an adapter-authoring contract.

Hard PreToolUse gate on AskUserQuestion that blocks questions whose decision context is invisible to the user. Three deterministic gates: deictic references to unseen context ('with that context', 'see above'), identical re-asks after the user pushed back in free text, and mid-workstream questions whose payload (question text, option descriptions, previews) and turn text both carry too little context. Block messages teach the exact recovery (put the context in the payload, then re-call). Fail-open on transcript errors; CCGM_ASK_CONTEXT_OFF=1 escape hatch.

Spec-driven development where E2E vision specs define target behavior. /atdd reads Playwright specs, iteratively builds app code until all tests pass, then ships.

Captures permission requests, tool failures (with error text and class), interrupts, and user-correction signals, plus a daily per-tool call counter; aggregates recurring failure signatures with plain code, then drafts a small rule_insert against real rule files through a direct Anthropic API call (no call when nothing qualifies); keeps every proposal in one ledger with states, tells the next session about ready fixes and a dead job (one SessionStart line a day), lets /autoheal-review accept a fix in one answer (a merged PR from a temporary worktree), and produces a local digest + opt-in Resend email; opt-in real-time security alerts and opt-in earned auto-apply (off, shadow, active after 10 reviewed decisions at 90% agreement), with each applied fix measured 14 days after merge and harmful ones reverted. Cross-clone-safe via fcntl file locks. Webhook publisher seam is dormant by default.

Configure Claude as a fully autonomous Staff-level engineer who executes tasks end-to-end without asking unnecessary questions.

/brainstorm - hard gate that forbids code, scaffolding, or implementation until a design spec is written and user-approved. Proposes 2-3 approaches with tradeoffs, writes a design doc, self-reviews for TBDs and contradictions, then hands off to /xplan. Pairs with /ideate (which refines the concept) to enforce spec-before-plan-before-code.

Hard PreToolUse gate that blocks Edit/Write/NotebookEdit and mutating git commands (commit, add, stage, apply) while HEAD is on the repo's default branch. Fires before the first edit so no work is ever produced on main. Bypass-proof (exit 2); ALLOW_MAIN_COMMIT=1 escape hatch; exempts in-progress rebase/merge/cherry-pick, unborn HEAD, and direct-to-main allowlisted repos.

Slash commands for brand naming research: /brand (full naming pipeline with word exploration, domain/trademark/app store checks) and /brand-check (single-name deep verification). Optionally adds Instant Domain Search MCP server for fast domain availability checks.

Rules for browser automation tool selection: Chrome extension, Playwright, and WebMCP. Includes verification priority order and UI verification workflow.

/capabilities - a decision map for CCGM's overlapping command/skill clusters. Answers "which one do I use?" for research, review, planning/execution, debugging, and knowledge/memory. A tight always-on rule points at the on-demand command so the full map costs zero idle tokens.

Audit tool for Claude Code installs. Three subcommands: check-resolvable (hook refs, command descriptions, script refs), dry (lexical overlap between command descriptions), and resolver-eval (runs a routing suite of {intent, expected} assertions against keyword-overlap scoring). Ships a default routing suite covering common slash commands.

Unified code-review orchestrator. Composes scope-drift, learnings-researcher, tiered reviewer personas (correctness, testing, maintainability, project-standards + conditional security, performance, reliability, api-contract, data-migrations), and an adversarial/red-team lens that runs AFTER the specialists with access to their findings. Confidence-gated autofix routing with safe_auto / gated_auto / manual / advisory classes. Modes - interactive / autofix / report-only / headless.

Delegate Claude Code agent work to Hetzner Cloud VMs. Provides golden image pipeline, VM lifecycle management, secret injection, workspace provisioning, and /dispatch command for one-command multi-agent dispatch.

Cloudflare-specific rules: Pages vs Workers selection, API-first Git integration, Email Service, and token scopes.

Code standards, testing requirements, error handling patterns, security practices, build verification, and living documents maintenance.

Essential slash commands: /commit, /pr, /cpm (commit-PR-merge), /gs (git status), /ghi (create issue).

Additional slash commands: /audit (codebase audit), /pwv (Playwright visual verify), /walkthrough (step-by-step guide), /promote-rule (promote repo rules to global), /freeze + /unfreeze + /guard (safety-hook state management), /checkpoint (save/resume WIP session state).

Experimental UserPromptSubmit hook that injects a compact preamble of iron-law principles (Confusion Protocol, Completeness, Evidence Before Claims, Root Cause Before Fix) at the start of slash-command invocations. Opt-in, disabled by default.

Utility commands and scripts: /cws-submit, /ccgm-sync (reverse-sync local config to CCGM), /user-test, statusline, agent-team launcher.

8 battle-tested anti-patterns to avoid: shallow directory exploration, dependency blindness, ESLint Fast Refresh, and more.

Team-shared learnings in docs/solutions/. After solving a non-trivial problem, /compound writes a structured markdown doc with YAML frontmatter that later /xplan and /review runs re-inject as grounding. Counterpart to self-improving's personal MEMORY.md.

/copycat - analyze external Claude Code config repos to find useful patterns, rules, and techniques worth adopting into CCGM

Opt-in restricted Claude/Codex review with frozen evidence, bounded exchanges, reusable evidence-bound acknowledgments and reversible Codex skill setup. Personal lead review and repository checks are the default.

/debug - structured root-cause debugging with Opus delegation. 7-phase workflow: gather context, reproduce, hypothesize, instrument, diagnose, fix, verify.

/deepresearch - Multi-query semantic research using the Exa MCP server. Claude generates diverse queries from your topic, fans them out via parallel Exa MCP tool calls, and synthesizes a structured research.md from full page contents. Requires the Exa MCP server registered via `claude mcp add` and an Exa API key.

/design-review - visual design review for web pages. Screenshots at 3 viewports, 6 parallel analysis passes: spacing, typography, responsive, visual hierarchy, accessibility, component consistency. Scored report with optional auto-fix.

Rule and template for shipping machine-readable docs alongside human docs. Any project an agent will install, build, test, deploy, or debug should have an AGENTS.md with copy-pasteable command blocks — not prose, not dashboard instructions.

/docupdate - comprehensive documentation audit and update. Checks README, TOC, onboarding flows, package lists, and module coverage against actual codebase state.

Seven-lens plan-quality gate. Before a plan, spec, or requirements doc ships to execution, /document-review fans out to 7 role-specific reviewer agents (coherence, feasibility, product-lens, scope-guardian, design-lens, security-lens, adversarial) and merges structured JSON findings with severity and confidence. Each lens has tight what-you-flag boundaries so they do not overlap.

Nightly, cost-capped dreaming service that mines Claude Code session transcripts for cross-session failure patterns and optimistically auto-integrates evidence-tagged memory-store changes (write-then-review, made safe by per-op-kind postures, a 24h dwell window, per-run blast caps, batch-anomaly detection, a windowed circuit breaker, an eval gate, a daily report, and git-backed rollback). Extends self-improving's learnings store with an out-of-band analyzer -- autoheal's capture-analyze-propose pipeline, retargeted at session transcripts instead of permission events. Provides the deterministic transcript miner (discover/mine/cluster/budget plus a schema-drift canary), the map-reduce analyzer (evidence bundle -> per-change proposals -> digest, with cost caps), the optimistic integration engine (per-op-kind postures, 24h dwell window, blast caps, batch-anomaly detection, windowed self-healing circuit breaker, plus an opt-in composite eligibility gate -- default off, add/supersede only -- scoring each mined add/supersede on four transcript-verified signals through a deterministic no-LLM waterfall before admission) with the /dream, /dream-digest, /dream-review, /dream-apply commands (no human queue: every proposal ends integrated or discarded with a reason, a once-a-day session line reports what integrated, /dream-review vetoes/reverts post-hoc, /dream-apply is a manual override; plus a one-off backlog close) plus the nightly LaunchAgent scheduler, and the memory eval harness (with/without-memory A/B plus a saturation third arm, a four-bucket outcome classifier, and an end-to-end task exercising the analyzer's own mined output -- establishing the --gate contract the optimistic engine consumes), and a read-only reconciliation report comparing Claude Code's own harness auto-memory against the learnings store (import candidates + contradictions, appended to the daily digest -- never writes to auto-memory).

/editorial-critique - deep editorial review of long-form writing. 8 parallel analysis passes: prose craft, AI-tell detection, argument architecture, conciseness, data verification, structure, impact, grammar. Judges prose against a lean, Hemingway-style house standard over ornate maximalism. Scored report with optional auto-fix.

Turns a topic into a narrated, Manim-rendered .mp4, rendered headless with deterministic checks (one video and one audio stream, duration within 10 percent of the narration, one frame per scene). Voice: ElevenLabs when a key is set, else macOS say. Opt-in rung 4 of the output ladder.

GitHub repository management protocols: issue-first workflow, PR conventions, label taxonomy, code review standards.

Git workflow rules: sync before history changes, rebase by default, post-merge cleanup, PR template detection, no AI attribution in commits.

Git worktrees as the default isolation for parallel sub-agent delegation on one machine, with a safe janitor that enforces teardown so worktrees never silently fill the disk.

Slim global CLAUDE.md that serves as the root configuration reference. Points to rules, commands, hooks, and settings without duplicating their content.

Python hooks that enforce git workflow rules: issue-first workflow, commit message format, branch protection, and auto-approval for file operations.

/ideate - structured ideation framework. Interviews you to refine a loose idea into a concrete concept using Socratic questioning, confidence tracking, and progressive refinement. Delegates to /deepresearch for validation and /xplan for planning.

Two foundational context files that give Claude Code a persistent identity: soul.md (AI personality and philosophy) and human-context.md (who you are, your goals, and how you work).

A /launch skill that takes a one-page spec and reaches a deployed Cloudflare Pages site without further human input — except a one-time Cloudflare GitHub App install, which the skill stops to ask for only if it's missing. Inspired by Karpathy's Sequoia talk: 'I would hope that I could give a prompt to an LLM, build menu gen, and then I didn't have to touch anything and it's deployed in that same way on the internet.' The skill is the forcing function that surfaces every place CCGM/CF infra is still human-shaped.

Keeps agent-driven live testing off the machine the operator works on. Anything that launches or relaunches an app, fires dictation, posts synthetic input events, changes focus, sets a machine-global input/audio override, or opens the mic runs only on the dedicated runner machine — and only against a permission grant recorded in the plan at planning time. Headless work (builds, unit tests, lint, read-only queries) stays on the dev machine.

Design-engineering details that compound into polished interfaces. Model-invoked skill with references on design direction (aesthetic identity, typeface/color/spacing choices, avoiding generic AI aesthetics), typography (text-wrap, tabular nums, font smoothing), surfaces (concentric radii, shadows over borders, optical alignment, image outlines), animations (interruptible transitions, enter/exit, icon micro-interactions), and performance (transition specificity, will-change).

Guide for building Model Context Protocol servers: project structure, tool design, error handling, testing with MCP Inspector, and evaluation patterns.

Security vetting gate for new AI models: before any open-weights or hosted model is integrated into the Claude Code harness or local dev system, verify weights provenance and integrity, file-format safety, license and data terms, the serving path, and stage its agentic access.

Multi-clone architecture for parallel agent work with issue claiming, port allocation, and the /mawf workflow command.

/onboarding - analyzes a repository and generates a structured ONBOARDING.md (architecture, dev setup, key commands, test workflow, glossary). Runs an inventory script to build a language-aware structural map, reads only the files that map surfaces, and writes prose with strict voice rules so the output reads like a knowledgeable teammate rather than generated documentation.

/orrery - deep-dives any codebase with parallel read-only scout agents and generates an interactive, zoomable, embeddable system-design map as one self-contained HTML file: 4 zoom tiers (landscape to key files), file-level GitHub links pinned to an anchor SHA, external systems, and per-node product-context prose. /orrery update refreshes an existing map incrementally. Ships the lockfile-pinned LikeC4 toolchain, the restricted orrery-scout agent, deterministic merge/emit/validate/render scripts, and publish-safety screening (secret quarantine, unconditional HTML escaping, sandboxed-iframe embed guide).

Output formatting for text addressed to the operator: copy-pasteable content goes in fenced code blocks, and a medium ladder (prose, diagram, page, video) picks the lowest medium that carries the content. Ships /explain and ccgm-md-page, a markdown-to-HTML page renderer for installs without the Artifact tool.

Packages the always-on TONE rules (terse communication, autonomous execution, copy-paste formatting) as Claude Code output styles — a system-prompt layer fixed at session start and prompt-cached — instead of always-loaded rules/*.md. Ships CCGM Terse plus CCGM ADHD (action-first output shaped for an ADHD reader). Opt-in alternative; does not remove the source rules.

Opt-in Stop hook that keeps the main agent working on a declared task until it is marked done. /persist starts, finishes, or cancels the loop; the hook blocks a stop only while a fresh, active, same-session state file exists, and fails open on context-limit stops, high transcript context, user aborts, auth errors, stale state, the iteration cap, a cancel signal, another session or project, re-entrancy, and any parse or IO error.

Maintainer tooling that projects CCGM's modules into a native Claude Code plugin marketplace (.claude-plugin/marketplace.json + per-module plugin.json), additively. The bash installer stays the canonical full-fidelity path. Ships the generator, a JSON-schema validator, and the SessionStart rule-injection hook that bridges the plugin-CLAUDE.md-not-loaded gap.

Structured resolver for PR review comments. /resolve-pr-feedback fetches unresolved review threads via GraphQL, triages new vs already-handled, and if 3+ new items arrive (or a cross-invocation signal fires) runs cluster analysis - categorizes each into 11 fixed concern categories and groups by category + spatial proximity. Clusters surface systemic issues instead of dispatching 10 one-off fixes. Parallel pr-comment-resolver subagents apply unambiguous fixes, post inline replies via gh api, and resolve threads; taste questions are batched for human decision.

Augments the external pr-review-toolkit plugin with scope-drift detection (compare stated intent to actual diff) and a Fix-First output format (AUTO-FIXED vs NEEDS INPUT). Ported from garrytan/gstack review skill.

Opt-in, backward-compatible relevance-scoped rule injection plus a tiered always-on safety core. Off by default: with the flag unset, all rules load exactly as before. When enabled, surfaces a pointer to the task-relevant rule subset while always including the safety core.

SSH access to a remote server. Adds /onremote command for health checks and remote command execution. Delegates to Haiku.

/research - multi-channel research using parallel agents with WebSearch, WebFetch, GitHub, and Reddit. Zero external dependencies. For deeper research backed by Exa neural search, see the deepresearch module.

Discipline for writing rules that hold up under pressure. Treats rule authoring as TDD: pressure-test a candidate rule with adversarial scenarios, capture the rationalizations agents use to bypass it, and rewrite until the rule closes those loopholes.

Meta-learning system with a structured JSONL learnings store (confidence decay, staleness, prompt-injection-filtered), automated reflection triggers, and hooks that fire at key moments (PR merge, context compaction, debugging failures). Replaces narrative MEMORY.md with a queryable, token-budgeted store while keeping MEMORY.md as a rendered index.

Cross-platform session historian agent. Searches prior Claude Code and Codex session transcripts for what was tried, what failed, and what was decided in earlier sessions on the same repo. Invoked by other skills (/compound, /xplan, /debug) to surface institutional knowledge a fresh session cannot see.

Structured session shutdown via /sds. Autonomous wrap-up sweep: commit dirty work, update referenced issues, run /reflect, write a handoff, broadcast to sibling clones. Complements /startup at the other end of the session.

Base settings.json with comprehensive tool permissions (800+ allow entries), deny list for dangerous operations, and plugin configuration. Defaults to 'ask' mode for safety.

shadcn/ui component patterns: composition over custom builds, semantic theming tokens, proper form architecture, accessibility, and CLI workflow.

/ship-ready command: at-a-glance dashboard of what gates a merge on the current branch (failing tests, open PR count, stale branches, outdated deps, merge velocity, review freshness, unresolved risks from docs/solutions/). Reads ce-review envelopes for commit-hash staleness detection.

Discipline for writing skills and slash commands that stay efficient, portable, and context-safe. Covers reference-file inclusion, conditional content extraction, tool selection, and writing style.

Slash command that promotes an ad-hoc session capability into a durable skill: a command file with triggers and rules, optional deterministic helper script, a pinning test, and a learnings-store entry pointing at the new skill. Inspired by the 'skillify' pattern where every repeated failure becomes structurally unreachable by being turned into a tested skill.

Plain-text /startup dashboard for Claude Code sessions. Surfaces git state, live sessions, tracking.csv claims, and recent activity (via session-history /recall). No agent-log writes; session transcripts are captured by Claude Code natively.

Compact single-line Claude Code statusline: model + effort, multi-clone identity (.env.clone), dir + git branch, context-used % with a compaction warning, session cost, and 5h/7d rate-limit bars. Installs the script and wires the statusLine setting.

Subagent dispatch methodology: task decomposition, spec-driven delegation, pass-paths-not-contents, two-stage review (spec compliance then code quality), parallel coordination, concurrency/rate-limit throttling for fan-outs, and skill invocation modes.

Supabase-specific rules: API key terminology (publishable/secret), environment variable naming, migration validation, and database change workflow.

4-phase root cause investigation methodology: investigate, analyze patterns, test hypotheses, implement fix. Prevents random fix attempts.

Tailwind CSS v4 design system patterns: CSS-first configuration, design token hierarchy, CVA component variants, dark mode, and responsive grid systems.

Strict red-green-refactor TDD discipline. No production code without a failing test first. Enforces test-first methodology for all new features and bug fixes.

Vision-driven e2e test suite generation. /test-vision for full repo analysis + parallel test suite creation. /e2e for single-feature spec generation. Composes /e2e as the atomic unit within /test-vision orchestration.

File-based review-finding tracker. Review findings, PR nitpicks, and tech debt that does not merit a full GitHub issue go to .claude/todos/NNN-{status}-{priority}-{description}.md with YAML frontmatter. Three skills compose - /todo-create (canonical writer), /todo-triage (interactive pending->ready), /todo-resolve (batch-resolve ready todos via parallel subagents). Fills the gap between review-found-it and cut-an-issue with a lightweight, in-repo third option.

Evidence-before-claims methodology. Requires fresh execution of verification commands, reading full output, and confirming results before asserting completion.

Orwell's six writing rules (1946) as an always-on prose standard for docs, PR text, commit messages, reports, and copy, plus a /rewrite command that applies them to existing text. Adds measurable STE-lite limits and a stdlib linter (ccgm-prose-lint). Governs prose only, never code or technical terms.

Research, planning and execution with an upfront choice of 1–3 adversarial reviews, default one. Personal lead review and repository checks are the default; alternating Claude/Codex review is opt-in.

/transcript <url> grabs a YouTube transcript via yt-dlp AND dispatches a subagent to produce an opinionated implications doc against your project memory. Saves both files with matching slug+date so the pair correlates by name. --no-analysis flag for extraction only.