Coding & Refactoring Agent Skill

US/global macro time series from St. Louis Fed FRED — CPI, GDP, Fed funds, yields, M2, DXY, etc.

US SEC EDGAR filings — list 10-K/10-Q/8-K/Form 4/S-1, fetch filing text, parse insider Form 4 transactions.

Create or update a minimal SPEC.md through a short interview, then use it as the source of truth for complex features, refactors, or ambiguous requests. Use when the user asks for /spec, spec-driven development, or when SPEC.md is missing or stale before substantial work.

Decide how to delegate work to subagents or teams. Agent-agnostic: covers Droid, Claude Code, Hermes, and Codex.

Search on-disk file contents across managed workspace repos via ripgrep (or Python fallback). Use when hunting code strings, configs, or Terraform/Docker patterns — not for manifest metadata.

Use when the user is debugging a bug, tracing an error, or asking why something fails. Examples: "Why is X failing?", "Where does this error come from?", "Trace this bug"

Use when the user asks how code works, wants to understand architecture, trace execution flows, or explore unfamiliar parts of the codebase. Examples: "How does X work?", "What calls this function?", "Show me the auth flow"

Use when the user asks about GitNexus itself — available tools, how to query the knowledge graph, MCP resources, graph schema, or workflow reference. Examples: "What GitNexus tools are available?", "How do I use GitNexus?"

Use when the user wants to rename, extract, split, move, or restructure code safely. Examples: "Rename this function", "Extract this into a module", "Refactor this class", "Move this to a separate file"

Multi-persona code review with structured findings. Use when the user says /review, asks for a code review, wants to review a diff/PR/branch before merging, or needs a pre-submit quality check.

Core execution discipline for any non-trivial task — multi-step work, anything that modifies files or systems, debugging, or work whose correctness matters, even when the user does not ask for rigor. The latest load moment is observable — you are about to take your first mutating action (edit, write, state-changing command); loading earlier is better, and if you notice you are already past it, load now. Governs the task contract, action gating, scope containment, verification by execution, adversarial self-review, and honest completion claims. Do NOT load for single-step questions, pure explanation, or trivial edits; for delegating to subagents or reviewing agent output, also load delegation-and-review.

How to write skills, CLAUDE.md/AGENTS.md, fix logs, and handoff documents that a weaker or zero-context model can actually execute. Load when authoring or updating any instruction file, SKILL.md, project memory, or institutional-knowledge document — including when converting lessons from a session into durable files. Not for writing user-facing docs or code comments.

Generates and formats git commit messages following the Conventional Commits specification. Use when the user asks for help writing commit messages, formatting git commits, wants to write a commit msg for staged changes, needs to produce a changelog entry, or mentions conventional commits, semantic versioning, or version bumps. Analyzes diffs or change descriptions, selects the correct type/scope, writes an imperative-mood header under 72 characters, and composes optional body and footer sections including breaking-change notation.

Complete Bun.js ecosystem guidance for runtime APIs, file I/O, package management, testing, SQLite, and security; use proactively when setting up Bun projects, replacing Node.js APIs with Bun-native APIs, writing bun test suites, implementing Bun.serve services, using bun:sqlite with prepared statements, configuring workspaces and lockfiles, hardening shell and SQL boundaries, or optimizing Bun performance and migration workflows.

Extract clean markdown content from web pages using Defuddle CLI, removing clutter and navigation to save tokens. Use instead of WebFetch when the user provides a URL to read or analyze, for online documentation, articles, blog posts, or any standard web page.

Comprehensive toolkit for validating, linting, testing, and analyzing Helm charts and their rendered Kubernetes resources. Use this skill when working with Helm charts, validating templates, debugging chart issues, working with Custom Resource Definitions (CRDs) that require documentation lookup, or checking Helm best practices.

Generate label matchers, line filters, log aggregations, and metric queries in LogQL (Loki Query Language) following current standards and conventions. Use this skill when creating new LogQL queries, implementing log analysis dashboards, alerting rules, or troubleshooting with Loki.

Comprehensive guide for creating software diagrams using Mermaid syntax. Use when users need to create, visualize, or document software through diagrams including class diagrams (domain modeling, object-oriented design), sequence diagrams (application flows, API interactions, code execution), flowcharts (processes, algorithms, user journeys), entity relationship diagrams (database schemas), C4 architecture diagrams (system context, containers, components), state diagrams, git graphs, pie charts, gantt charts, or any other diagram type. Triggers include requests to "diagram", "visualize", "model", "map out", "show the flow", or when explaining system architecture, database design, code structure, or user/application flows.

Integrate Bun runtime into Nx monorepos with deterministic plugin setup, executor configuration, migration from Node.js toolchains, and cache-aware build/test workflows; use when adding the nx-bun plugin, converting projects, or standardizing Bun targets across Nx workspaces.

Build, install, initdb, and start PostgreSQL from source in the `dev/` clone for backend hacking — covers meson setup (PG ≥ 16 default) with cassert + debug flags, the autoconf ./configure fallback, ninja install, initdb + pg_ctl start / stop, PGDATA / PATH wiring, single-user mode for postmaster startup debugging, attaching gdb / lldb under the per-connection fork model, and -O0 -g3 debug builds. Use whenever a task involves compiling PG from source in dev/, running ninja install on the dev clone, initdb-ing a fresh data directory, starting or stopping the dev cluster via pg_ctl, picking between the debug profile (5432) and ASan profile (5433), or attaching a debugger to a forked backend. Skip for brew / apt / yum / Docker / k8s installation of release PG, Aurora / Cloud SQL / Supabase / Neon-managed PG provisioning, generic CMake / make / Bazel build questions, Linux-kernel builds, Node.js / Python / Go application builds, and pgAdmin / DBeaver client installation.

Format C code to upstream PostgreSQL house style for src/backend / src/include — covers hard tabs at width 4, BSD braces, postgres.h-first include order, C99 subset rules (no // comments, no VLA, no mid-block declarations), naming conventions (struct typedef + field naming, function names matching typedef names), function-header comment format, ~78-char line length, and pgindent expectations. Use whenever a PG patch edits, adds, or reviews .c / .h files under source/src/ or dev/src/, or when a reviewer flags pgindent churn on a posted patch. Skip for Linux-kernel style (CodingStyle), clang-format / rustfmt / prettier / black / shfmt configurations, non-PG C / C++ style (Google style, LLVM style, Mozilla style), Java checkstyle, JavaScript ESLint, EditorConfig tuning, and general "what's a good C style" advice.

Traverse the pg-claude knowledge graph to answer "what does this feature/file/pattern touch across the corpus?" — pulls scenarios, idioms, call-site file examples, sibling patterns, subsystem ownership, and analogous past features from planning/ + sessions/ into a single chain map. Uses the graph edges built by `scripts/populate-idiom-callsites.py` (idiom → files) and `scripts/build-scenario-idiom-matrix.py` (scenario ↔ idiom bidirectional). Use proactively when brainstorming a new PG feature, planning §3 file table, investigating an unfamiliar subsystem, or trying to find "have we touched something like this before?". Also use inside `pg-feature-brainstorm` step 1 (subsystem framing) and `pg-feature-plan` before §3 to seed the file list from existing evidence rather than pure grep. Skip when you already have the anchor's downstream chain memorized, when the task is a one-file bug fix, or when the ask is about non-PG code.

Write or review a PostgreSQL backend ereport / elog call — covers ereport vs elog, picking a SQLSTATE from errcodes.txt, errcode_for_file_access, errmsg / errdetail / errhint capitalisation rules, soft errors via escontext, PG_TRY / PG_CATCH longjmp-safe cleanup blocks, and the DEBUG / LOG / NOTICE / WARNING / ERROR / FATAL / PANIC elevel ladder. Use whenever a PG patch adds, edits, or reviews C in source/src/backend or contrib/ that reports an error or logs a message — picking elevel, choosing a SQLSTATE, formatting errmsg, wiring PG_TRY/PG_CATCH around a longjmp-unsafe block, or migrating a call to soft-error style. Skip for Python try/except, Go error returns, Rust Result / anyhow / thiserror, C++ exceptions, Java checked exceptions, Sentry / pino / Winston / log4j application logging, Oracle ORA-* / MySQL error codes, and general error-handling philosophy questions.

PostgreSQL's MultiXact machinery — `src/backend/access/transam/multixact.c` — the multi-transaction ID system that lets a single heap tuple record multiple concurrent lockers (for `SELECT ... FOR SHARE` and `SELECT ... FOR UPDATE` blends) plus the multixact-freeze / wraparound path. Loads when the user asks about MultiXactId semantics, `xmax` interpretation with the `HEAP_XMAX_IS_MULTI` bit, why a tuple can appear to be locked by multiple transactions, tuple-locking modes (`FOR SHARE` / `FOR NO KEY UPDATE` / `FOR KEY SHARE`), the multixact SLRU (`pg_multixact/`), MultiXact member offsets, or the `autovacuum_multixact_freeze_max_age` wraparound path. Skip when the ask is about clog (regular transaction commit status — sibling but different code) or 2PC prepared transactions.