forge-dev

Senior architecture reviewer — evaluates system designs, architectural decisions, scalability, integration, and evolution paths. USE WHEN reviewing an ADR, assessing a proposed architecture change, evaluating service boundaries, or challenging a technology choice.

Senior code reviewer covering correctness, security, performance, maintainability, and test quality across TypeScript, Python, Rust, Go, and SQL. USE WHEN reviewing a diff, before merging a PR/MR, assessing code quality, or surfacing security and performance risks in recent changes.

Senior error and log forensic analyst — correlates distributed failures, uncovers root causes, and maps cascades. USE WHEN diagnosing a production incident, correlating errors across services, analyzing cascading failures, or mining log patterns for hidden root causes.

Senior refactoring specialist — transforms complex code into maintainable systems while preserving behavior. USE WHEN refactoring legacy code, reducing complexity, breaking apart god classes, extracting methods, or planning a safe multi-step refactor.

Architecture for security and health audit scripts: standalone-runnable checks, severity ladder with UNKNOWN, key:value output contract, orchestrator dispatch, exit-code semantics. Language-agnostic; reference implementations in Bash, applies to Python and other languages. USE WHEN writing audit tools, designing check-script contracts, adding checks to tools like check-mac, reviewing audit-script architecture, or porting an audit tool between languages.

Bash and POSIX shell conventions for the forge ecosystem — idioms over subprocesses, fail-safe defaults, probe guarding, multi-line command substitution, exit-code semantics. USE WHEN writing or reviewing shell scripts (audit tools, hooks, build scripts), designing CLI flag handling, or wiring shell tools into CI.

Proactive code cleanup — duplication removal, pattern adoption, defensive hardening, readability maximization. USE WHEN cleanup, refactor, code quality, reduce duplication, adopt patterns, readability, make code explicit, clean up code, polish code.

Build static HTML data explorer apps from SQL databases or datasets. Vega-Lite or Chart.js for charts, Tabulator for tables, vendored JS. Produces standalone HTML files suitable for SharePoint, Cloudflare Pages, or email.

Choose and configure visualization tools for data explorer apps. Vega-Lite for declarative charts, Chart.js for Sankey and custom visualizations, Tabulator for interactive tables.

Defensive programming principles for robust systems — input validation, error boundaries, fail-safe defaults, invariant checking. USE WHEN writing production code, designing APIs, handling external input, or reviewing error handling.

Decide how to integrate completed work — merge, PR, or cleanup — once implementation is done and tests pass. USE WHEN implementation is complete and tests pass, deciding merge strategy, or cleaning up a feature branch.

Inspect failing GitHub Actions checks via gh, pull logs, summarize root cause, and propose a fix plan. USE WHEN a GitHub PR has failing CI checks, debugging GitHub Actions failures, or triaging CI breakage before merging.

Fix issues end-to-end — triage, diagnose, branch, fix, PR/MR, close. Supports GitHub (gh) and GitLab (glab) with cross-repo fixes and submodule bumps. USE WHEN fix issue, github issue, gitlab issue, bug report, fix bug, resolve issue, triage issue, gh issue, fix

Run tests and systematically fix all failing tests using smart error grouping. USE WHEN tests are failing, user asks to fix failing tests, CI reports test failures, or a test suite run leaves failures.

GitHub CLI reference — repo management, releases, Actions/CI, PR operations, branch protection, labels. USE WHEN github, gh, repo, release, actions, ci, workflow, branch protection, labels, milestones.

Process code review feedback with technical rigor — verify before implementing, push back when wrong. USE WHEN receiving code review feedback, before implementing review suggestions.

Author a clear scoped request when asking for a code review. USE WHEN ready for review, finishing a task, opening a PR, or requesting review feedback on a piece of code.

Rust development conventions — crate structure, error handling, CLI patterns, config loading, testing, and cross-platform compilation. USE WHEN writing Rust code, creating binaries, designing library APIs, or reviewing Rust implementations.

Delegate independent tasks from an implementation plan to subagents. USE WHEN a plan has independent tasks suitable for parallel subagent execution, multi-agent implementation, or delegating parts of a plan.

Four-phase debugging methodology — root cause before fixes. USE WHEN encountering any bug, test failure, unexpected behavior, or build failure.

Test-driven development practices — Red-Green-Refactor cycle, test categories, coverage strategy, property-based testing. USE WHEN writing tests, designing testable APIs, or reviewing test coverage.