mz-dev-pipe

Pipeline-only. Reviews implementation code for bugs, security issues, missed requirements, and convention violations by reading every modified file thoroughly.

Pipeline-only. Implements specific work units from an approved plan. Reads existing code, follows project conventions, and makes precise changes.

Pipeline-only. Final quality gate. Verifies the task is 100% complete by checking implementation against requirements, plan, and all review feedback. Can trigger pipeline restart from any phase.

Pipeline-only executor agent dispatched by skill orchestrators. Runs lint and/or format commands, parses per-file issues, and writes a structured lint_results.md artifact. Missing or unconfigured tools are noted but never block. Never user-triggered.

Pipeline-only. Cleans and optimizes code after implementation is functionally complete. Removes dead code, debug artifacts, unused imports, duplicated logic, and unnecessary complexity without changing behavior.

Pipeline-only. Critically reviews implementation plans. Catches missing integration points, incorrect parallelism, incomplete test strategies, and architectural issues.

Pipeline-only. Creates detailed, parallelizable implementation plans. Breaks tasks into independent work units with clear file paths, changes, and dependencies.

Pipeline-only. Explores codebases and researches domains. Gathers context about project structure, patterns, conventions, and external domain knowledge needed for implementation planning.

Pipeline-only. Reviews test coverage completeness. Identifies untested functions, missing code paths, and gaps in edge case coverage.

Pipeline-only. Reviews test quality. Evaluates whether tests are meaningful, independent, well-structured, and would catch real regressions.

Pipeline-only executor agent dispatched by skill orchestrators. Runs the project test suite (or a subset), parses per-test results, and writes a structured test_results.md artifact. Handles scoped runs and specific test file lists. Never user-triggered.

Pipeline-only. Writes comprehensive tests for implementations. Creates unit tests, edge case tests, and integration tests following project conventions.

Pipeline-only detector agent dispatched by skill orchestrators. Reads project manifests to detect test, lint, format, and type-check tooling and writes a structured tooling.md artifact. Returns cached `.mz/tooling.json` when present unless `force_refresh: true` is specified. Never user-triggered.

Pipeline-only. Translates one file or chunk with Bash-verified placeholder and structural fidelity. Preserves markdown, fenced code, and i18n tokens. Emits a per-chunk confidence report; supports lean back-translate round-trips.

Pipeline-only. Web-first research agent for pipeline skills. Conducts multi-source research, cross-references primary sources, and produces structured findings. Used by /combine for gap-fill and any pipeline skill that needs external verification.

ALWAYS invoke when the user wants to translate files, docs, or i18n resources. Triggers - "translate X to Y", "localize", "i18n to <lang>". When NOT to use - single-sentence phrase translation (answer inline).

Orchestrates a deep verification pass that checks whether code in scope is correct, clean, well-tested, and functional. This pipeline reports findings — it does NOT auto-fix anything. The user decides what to fix based on the report.

Orchestrates a multi-lens codebase audit. Parallel researchers scan across correctness, security, performance, maintainability, and reliability lenses. Findings are consolidated, ranked, approved by the user, then fixed in parallel with mirrored reviewers. Critical/high fixes get regression tests.

ALWAYS invoke when the user wants to see the impact of changing a file, function, or module before refactoring. Triggers: "blast radius of X", "what depends on X", "impact analysis". When NOT to use: the refactor itself (use optimize or build).

ALWAYS invoke when the user wants to build, implement, or create a new feature, module, or component from scratch. Triggers: "build X", "implement Y", "add feature". When NOT to use: bug fixes (use debug), polishing existing code (use polish).

You orchestrate a local-first synthesis that harvests prior pipeline output — `.mz/research/`, `.mz/task/*/`, `.mz/reports/`, `.mz/reviews/`, codebase files, and git history — and folds it into a unified report with task-adaptive sections (or user-supplied sections). Counterpart to `/deep-research`: that skill is web-first and starts from zero; this one assumes prior output exists and only consults the web to fill residual gaps behind an explicit approval gate.

Orchestrates a reactive bug investigation: reproduce the bug, diagnose root cause (with optional domain research), get user approval on the diagnosis, write a regression test that fails (TDD), fix minimally, verify, review, and report.

You orchestrate a deep-dive code analysis that produces a comprehensive, diagram-rich report explaining how a section of code works, why it's designed that way, and what potential issues exist. This is a read-only pipeline — it never modifies source code.

You orchestrate a hypothesis-driven investigation: analyze the codebase against a suspected issue, run domain research when the hypothesis involves complex external behavior, write exploratory tests to prove or disprove the hypothesis, and report findings with a verdict. No code fixes — output is a report only.

Orchestrates a multi-agent optimization pass over existing code. Builds import-graph-based chunking, dispatches parallel `pipeline-optimizer` agents with mirrored `pipeline-code-reviewer` agents per chunk. On rejection, respawn rejected chunks only. Tests and linters run after each batch.

Orchestrates iterative polish of existing code against specific completion criteria. Unlike `build` which builds from scratch, polish works with what's already there — running tests, diagnosing failures, fixing issues with review loops, and optimizing.

2 hooks across 2 events