beigebox

Use when the user wants to fan-out / distribute / split / parallelize a list of inputs across multiple model calls so a single oversized prompt doesn't blow the context or token budget — especially relevant for reasoning models (trinity-large-thinking, deepseek-r1, o3) whose internal reasoning consumes the budget before any visible output. Takes a list of items, a per-item prompt template, and a model; runs N parallel calls against the BeigeBox proxy with a configurable concurrency cap; optionally reduces the responses with a final merge call. Item sources are file (one per line), JSON list, glob (each match becomes {path, contents}), or stdin.

Use when the user wants to fuzz / find DOS or crash bugs in / dynamic-analyze a Python repository, or wants the dynamic complement to a static-analysis pass (garlicpress, Trinity). Walks the repo, picks the highest-risk fuzzable functions, runs each in an isolated subprocess harness with a mutation loop seeded from docstrings + parser edge cases, and emits garlicpress-shape Finding dicts. Python-only; output is JSON or a one-line-per-finding summary.

Use when the user wants static analysis / SAST / type-checking / lint-with-teeth on a Python codebase — finding security smells, logic bugs, type errors, obvious defects without running the code. Wraps ruff (full ruleset including the bandit-port S rules), semgrep (registry-backed pattern + dataflow rules), and mypy (type checking). Emits garlicpress-shape findings so the output merges with fuzz/static pipelines without translation. Three categories, not duplicates: ruff covers AST patterns + bandit; semgrep covers cross-file dataflow; mypy covers types. Skip bandit (ruff -S already covers it).

Use when the user wants to audit / profile / fingerprint one or more machines to learn OS, hardware, network interfaces, installed virtualization stacks, running containers/VMs, and listening services. Takes a list of targets — local, SSH alias, user@host, or user:password@host — and works against Linux and macOS regardless of authentication style. Produces a markdown report (default), JSON, or a CLAUDE.md-ready section.

Use when the user asks what services/containers/VMs are running, how to connect to something (URL, IP, shell), or wants an inventory of endpoints across local and remote hosts. Works on Linux and macOS. Probes Docker, Podman, Incus, LXD, classic LXC, libvirt, systemd-nspawn, OrbStack, Colima, Multipass, VirtualBox, Parallels, VMware Fusion, and Tart. Supports SSH to remote hosts and emits either a human table or JSON.