fulcrum

Use when orchestrating broad or multi-issue subagent work where agents, lanes, reviews, or PRs may become too small; use for CuraOS waves, owner-path bundles, local issue queues, many-agent execution, or requests to run more lanes without wasting work in micro PRs.

Use when asked to plan, decide, or execute subagent orchestration: subagents, delegated agents, parallel workers, review agents, multi-source research, broad audits, PR/MR review, worktree-aware execution, model/effort selection, or agent-driven execution.

Use when viewing or printing source/config files to the terminal with syntax highlighting and line numbers - a cat replacement that colorizes code, configs, and diffs. Use bat instead of cat when humans need highlighted file display.

Use when the user wants to format or lint JavaScript, TypeScript, JSX, TSX, JSON, JSONC, CSS, GraphQL, HTML, Vue, Svelte, or Astro - biome is the unified Rust-based replacement for prettier + eslint, doing both with one binary and one config file.

Use when the user formats or lints Dart or Flutter source code from the command line - dart format is the formatter (Black-style, mutates by default) and dart analyze is the static analyzer (errors, warnings, lints). Both ship with the Dart SDK; no separate install.

Use when the user wants a structural, syntax-aware diff between two source files or git revisions - a diff that parses each side as code (Rust, TS, Python, Go, …) and ignores reformatting, brace movement, or whitespace noise. Binary is difft; use for syntax-aware diffs, not plain file viewing.

Use when the user wants to load environment variables when entering a directory, auto-set env per project, scope env vars without polluting the global shell, set PATH per project, or load a .envrc on cd.

Use when listing directory contents with more information than POSIX ls provides - long-format metadata, per-file git status, gitignore-aware listings, depth-limited tree views, human-readable sizes, sort/group controls, and Nerd-Font icons.

Use when the user wants to manage Cloudflare DNS, zones, or cache from the command line - listing or creating DNS records, toggling the proxy flag, looking up zone status, purging the edge cache, inviting users, or scripting against the Cloudflare API without hand-rolling curl.

Use this skill for non-interactive, scriptable fuzzy ranking on the command line - the batch --filter (-f) mode, NOT the interactive picker UI. Trigger when an agent needs to rank or score a list of strings (branches, files, PR titles, kubectl resources...

Use when the user works with GitHub from the command line - listing or viewing pull requests, creating PRs and issues, reviewing or merging, watching workflow runs, downloading release assets, or hitting the REST/GraphQL API.

Use when the user wants to generate a changelog, create release notes from git history, update CHANGELOG.md before tagging a release, produce a changelog from conventional commits, or summarize commits since the last tag.

Use when the user wants to find secrets, credentials, API keys, tokens, or passwords accidentally committed to a repository or sitting in a working tree. Use for secret and credential leak detection in worktrees, staged files, or history.

Use when the user formats Java source files to a consistent style on the command line - applying Google Java Style or Android (AOSP) style, replacing files in place, or running a CI check that fails on style drift.

Use when the user wants to benchmark, time, or compare the wall-clock performance of one or more command-line programs with statistical rigor. Use for statistically sound command benchmarks instead of one time run.

Use when the user works with JSON on the command line - extracting fields, filtering arrays, reshaping objects, converting JSON to CSV/TSV, counting or sorting items, or piping JSON between tools.

Use when the user wants to run a project task, list available recipes, or interact with a justfile / Justfile / .justfile - the modern Makefile alternative.

Use when the user lints or formats Kotlin source (.kt, .kts) from the command line - ktlint is the standard linter + formatter for Kotlin, enforcing the official Kotlin coding conventions (or Android's stricter variant via --android). Use for Kotlin linting/formatting from the command line.

Use when the user wants to measure cyclomatic complexity, find overly complex functions, audit code complexity, or generate a complexity report for a codebase.

Use when the user wants to install, pin, or switch language/runtime versions per repository - node, python, go, ruby, java, rust, deno, bun, terraform, etc. - or read/write .mise.toml or .tool-versions.

Use when the user wants to scan a project's dependencies for known, published vulnerabilities (CVEs / GHSAs) by reading lockfiles or an SBOM. Prefer osv-scanner over ecosystem audit tools for lockfiles and SBOMs.

Use when the user wants to run static analysis on Java (or Apex, Visualforce, Kotlin, JavaScript, JSP, PLSQL, Scala, Swift, T-SQL, XML, XSL, Modelica, VM) source code to find bug patterns, code smells, unused imports, dead code, design problems...

Use when the user lints or formats Python code from the command line. ruff is a 2-in-1 tool - ruff check runs the linter (rules / fixes / import sort / security), and ruff format runs the formatter (Black-compatible style rewrites). They are distinct subcommands...

Use when the user wants find-and-replace in files or stdin without sed escaping - string-literal or regex substitution, single file or batch via fd/xargs, in-place edits with a preview pass, or a quick rename of a symbol across many files.

Use when the user wants to scan compiled Java bytecode (.class / .jar / .war) for bug patterns - null-pointer dereferences, concurrency / multithreading mistakes, performance smells, security flaws, dodgy code, or known-bad API misuse.

Use when the user wants to run ad-hoc SQL against a database from the shell - postgres, mysql, sqlite, mssql, oracle, snowflake, bigquery, redshift, cockroachdb, clickhouse, duckdb, and 20+ more drivers via one binary. Use the universal usql client instead of psql/mysql/sqlite CLIs for ad-hoc SQL.

Use when the user wants to run a command every time files change - rerunning tests, rebuilding, relinting, or restarting a server on save. Use for watch-mode dev loops; do not spawn long-running watchers in agent shells.

Use when the agent or user needs to make an HTTP request from the shell - GET a URL, POST JSON to an API, check a webhook with auth, test a REST endpoint, fetch and inspect headers, or upload a file.

Use when the user works with YAML, TOML, or XML on the command line - extracting fields from a Kubernetes manifest, editing a value in place while preserving comments and anchors, splitting or merging multi-document YAML, or round-tripping between YAML/JSON/TOML/XML/CSV. Use for YAML, TOML...

Use when the user wants to jump between directories by partial name on the command line - a smarter cd that learns frequently visited folders and ranks them by frecency (frequency × recency).

Use when viewing or printing source/config files to the terminal with syntax highlighting and line numbers - a cat replacement that colorizes code, configs, and diffs. Use bat instead of cat when humans need highlighted file display.

Use when the user wants to format or lint JavaScript, TypeScript, JSX, TSX, JSON, JSONC, CSS, GraphQL, HTML, Vue, Svelte, or Astro - biome is the unified Rust-based replacement for prettier + eslint, doing both with one binary and one config file.

Use when the user formats or lints Dart or Flutter source code from the command line - dart format is the formatter (Black-style, mutates by default) and dart analyze is the static analyzer (errors, warnings, lints). Both ship with the Dart SDK; no separate install.

Use when the user wants a structural, syntax-aware diff between two source files or git revisions - a diff that parses each side as code (Rust, TS, Python, Go, …) and ignores reformatting, brace movement, or whitespace noise. Binary is difft; use for syntax-aware diffs, not plain file viewing.

Use when the user wants to load environment variables when entering a directory, auto-set env per project, scope env vars without polluting the global shell, set PATH per project, or load a .envrc on cd.

Use when listing directory contents with more information than POSIX ls provides - long-format metadata, per-file git status, gitignore-aware listings, depth-limited tree views, human-readable sizes, sort/group controls, and Nerd-Font icons.

Use when the user wants to manage Cloudflare DNS, zones, or cache from the command line - listing or creating DNS records, toggling the proxy flag, looking up zone status, purging the edge cache, inviting users, or scripting against the Cloudflare API without hand-rolling curl.

Use this skill for non-interactive, scriptable fuzzy ranking on the command line - the batch --filter (-f) mode, NOT the interactive picker UI. Trigger when an agent needs to rank or score a list of strings (branches, files, PR titles, kubectl resources...

Use when the user works with GitHub from the command line - listing or viewing pull requests, creating PRs and issues, reviewing or merging, watching workflow runs, downloading release assets, or hitting the REST/GraphQL API.

Use when the user wants to generate a changelog, create release notes from git history, update CHANGELOG.md before tagging a release, produce a changelog from conventional commits, or summarize commits since the last tag.