claude

ccmaf

CCMAF v2 core: the full framework as a plugin — board/lifecycle machinery, /ccmaf:init scaffolding, fat-plan/thin-build workflow, doctor, per-task budgets, unattended safety profile. Requires ccmaf-kernel; activates only on v2-marked projects.

2 on GitHubv2.0.4MITupdated 1 month ago
/

Plugin ccmaf from marketplace ccmaf in drushegh/CCMAF. Commands read from the repo's manifest — 1 month ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Slash commands (16)

  • /ccmaf:analyse

    Architect analysis of requirements and existing code into a design and task breakdown (CCMAF v2)

  • /ccmaf:board-heal

    Reconcile the TASKS.md board with reality — advance stranded tasks, backfill verify seeds, fix IDs (CCMAF v2)

  • /ccmaf:bug

    Log an ad-hoc bug into the TASKS.md Bug-Fix Lane in the shared on-disk bug shape (CCMAF v2)

  • /ccmaf:build

    THIN build loop — dispatch against plan specs with budgets, receipts, integration gate, staged review (CCMAF v2)

  • /ccmaf:healthcheck

    Deep periodic audit across framework integrity, code quality, contracts, state files and tests (CCMAF v2)

  • /ccmaf:housekeeping

    Framework maintenance — rolling summary, state archiving and distillation, feedback reconciliation (CCMAF v2)

  • /ccmaf:init

    Scaffold this project onto the CCMAF v2 framework (state files, settings, aliases, feature opt-ins)

  • /ccmaf:plan

    FAT planning — contracts, 1:1 story tasks, per-task acceptance specs + budget classes (CCMAF v2)

  • /ccmaf:post-compact

    Rehydrate after a manual /compact by reconciling the lossy summary against on-disk state (CCMAF v2)

  • /ccmaf:pre-compact

    Flush the in-context working set to disk before a manual /compact (CCMAF v2)

  • /ccmaf:reconcile

    Reconcile changed code against the existing codebase and contracts - duplicates, seam mismatches, contract and convention drift (CCMAF v2)

  • /ccmaf:review

    Review Ready-for-Review tasks — receipts to the orchestrator, findings to disk, verified before acted on (CCMAF v2)

  • /ccmaf:security

    Security sweep — SAST, secret scanning, dependency scanning and license/provenance checks (CCMAF v2)

  • /ccmaf:test

    Test Ready-for-Test tasks against their plan specs; emit verify seeds; receipt returns (CCMAF v2)

  • /ccmaf:update

    Check plugin updates + apply CCMAF v2 scaffold deltas to this project

  • /ccmaf:wrapup

    End-of-session handoff — flush context to disk, commit and push before a restart or machine switch (CCMAF v2)

Subagents (9)

  • architect

    System architect for planning, contracts, and design decisions. Use when designing features, defining API contracts, making technology choices, or planning module structure. Never writes production code (developer's job), runs tests (tester), or reviews code (reviewer).

  • developer

    Implements features and fixes bugs in production code. Use whenever a task requires writing or modifying application source, tests, configuration, or build tooling. Not for system design (Architect), test plans (Tester), or code review (Reviewer).

  • reconciler

    Opus-tier horizontal auditor — reads the project sideways across modules and agents instead of vertically within one task. Detects semantic duplicates, cross-module/cross-agent seam mismatches the type system can't see, contract drift against ECOSYSTEM.md and ARCHITECTURE-style lists, and convention

  • researcher

    Evidence-gathering specialist — external research (docs, APIs, libraries, standards) and internal archaeology (git history, existing docs). Use whenever a decision needs facts nobody in the session can verify from the code alone — library choices, API behaviour, version compatibility, domain rules,

  • reviewer

    Code reviewer for contract compliance, security, and quality. Use whenever a task is Ready for Review — approval moves it to Ready for Test, issues found send it back to In Progress — when settings.json/hooks/agents change, or for periodic security sweeps. Returns structured findings only — never mo

  • tester

    Validates implementations against contracts and writes tests. Use whenever a task moves to Ready for Test, whenever a bug fix needs verification, or when contract changes need new test coverage. Does not write production code or fix bugs.

  • ui-designer

    Visual and interaction design implementer. Use whenever a task creates or meaningfully changes user-facing UI — new screens, components, styling passes, "make it look better" requests. Proposes design directions before building, works from explicit design tokens, and never hands off unrendered work

  • ux-critic

    Adversarial UX and visual-design critic. Use whenever user-facing UI is about to be accepted — after ui-designer or Developer work on screens, flows, or styling. Performs cognitive walkthroughs with numeric friction counts and rubric-scored visual critique; briefed to refute "the flow is fine". Retu

  • verifier

    Adversarial verifier of claims made by other agents — review findings, fix claims, test results. Use whenever another agent's output is about to be acted on (findings moving a task, a "fixed" claim closing a bug, a green suite gating a merge). Always a fresh context; never the agent that produced th

Hooks (12)

  • SessionStart

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/core-context.sh"'

  • SessionStart

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/session-reanchor.sh"'

  • SessionStart

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/session-start-marker.sh"'

  • SessionStart

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/guard-interpreter-check.sh"'

  • UserPromptSubmit

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/framework-drift-guard.sh"'

  • UserPromptSubmit

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/checkpoint-watermark.sh"'

  • PreToolUse

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/unattended-guard.sh"'

  • PostToolUse

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/checkpoint-watermark.sh"'

  • PostToolUse

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/task-budget-counter.sh"'

  • PreCompact

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/precompact-snapshot.sh"'

  • PostCompact

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/postcompact-archive.sh"'

  • Stop

    bash -c 'cd "$CLAUDE_PROJECT_DIR" && bash "${CLAUDE_PLUGIN_ROOT}/hooks/enforce-state-update.sh"'

View source on GitHub

More from drushegh/CCMAF

All 7 plugins →

advisors

claude

Multi-model advisory workbench: /advisors:fable /sol /terra /luna /consult /crossbench /mode — staged consultations with fail-closed model attestation.

7 commands

2

council

claude

Council of Agents: five adversarial advisor perspectives + anonymised peer review + a chairman verdict with transcript and HTML report.

1 command · 6 agents

2

console

claude

Claude-side bridge to the ccmaf-console web dashboard (npm): session autostart + heartbeat hooks and the launcher driver.

1 command · hooks

2

CCMAF Tier 0 kernel: dangerous-command guard, tool-call budget, done-criterion + receipt discipline. Zero per-project files.

hooks

2

devhooks

claude

Stack-aware dev QoL hooks for any repo: format-on-edit, lint-on-edit, dependency verification, noisy-test-output filtering.

hooks

2

media

claude

Media generation via the codex CLI ChatGPT-sub built-ins (zero metered spend): /media:image renders images through gpt-image.

1 command

2

Maintain ccmaf?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins7 plugins