Skip to content

reconFTW is a tool designed to perform automated recon on a target domain by running the best set of tools to perform scanning and finding out vulnerabilities

6.8 KB
Raw Blame

reconFTW v2 beta — v5.0.0-beta.1

reconFTW has been rewritten from bash to Go. This is the first public build of that rewrite, and it ships as an opt-in pre-release.

This document is the canonical announcement for v5.0.0-beta.1. It leads with what the beta is, and then — at more length, because it matters more — with what it is not.


What this is

A single statically-linked binary that replaces the sourced-bash engine. Same recon philosophy: one command, a complete passive/active/vulnerability picture of a target. Different machinery underneath:

  • Resumable by construction. Every task records a checkpoint keyed by a hash of its inputs — config slice, target, wordlist contents. Re-running skips completed work.
  • Structured output. Artefacts are JSONL with a schema, written atomically, scope-checked at the write boundary.
  • Bounded concurrency. One scheduler with a global ceiling and per-tool rate limits, instead of unbounded background jobs sharing global state.
  • Portable. No bash 4.3+ requirement, no GNU-coreutils-on-macOS dance. Linux and macOS, amd64 and arm64, plus a fully-static musl build for Alpine.
  • TOML config, with a migrator that reads your existing reconftw.cfg and tells you loudly about every key it could not map.

The ~100 external tools are still the ~100 external tools. The rewrite changed the orchestrator, not the toolbox.


What this is NOT

Three things are open. None of them is a surprise we are hoping you will not notice; they are the reason the beta exists at all.

1. Parity is measured against ONE target, not three to five

Bug-for-bug parity against bash v1 has been measured on one canonical target, run side-by-side with the outputs diffed. The acceptance criterion, CUT-11, asks for three to five canonical targets.

That gap matters in a specific way: differences on any other target are unmeasured, not verified-absent. If v2 returns something bash v1 did not — or misses something bash v1 found — on a target we have not run, no test currently in the tree would have caught it. Please report those. A side-by-side diff on a target we have never run is the single most valuable thing a beta tester can send us.

2. Throughput is unmeasured

No claim is made that v2 is faster than bash v1, or slower. Nobody has benchmarked the two end to end against the same targets under the same rate limits, so any number you might expect to see here would be invented.

The architecture changed in ways that plausibly cut either way — a bounded scheduler replaces unbounded background jobs, which can trade peak speed for predictability. Which way it actually lands is an open empirical question. If you run both, the timings are worth reporting; treat any assertion about v2's speed, including a favourable one, as unsupported until then.

3. Recon/<domain>/ is not created — CUT-08 is open

This is the one most likely to break something of yours.

The beta does not create the top-level Recon/<domain>/ tree. The bash-shape output tree is written under the workspace root as _compat/ instead. Filenames inside it follow the v1 shape; the parent path does not.

So a pipeline that reads Recon/<domain>/subdomains/subdomains.txt finds nothing under the beta, and a pipeline pointed at _compat/ finds what it expects.

Whether v2 should also create the top-level Recon/<domain>/ path is an open decision, tracked as CUT-08 — not an oversight, and not something the beta is quietly hoping to skip. It is open because the answer depends on how many people actually script against that path and how, which is exactly what a beta period is for. If this breaks a script of yours, say so. That report is the input the decision is waiting on.


Why the tag is v5.0.0-beta.1 and not v2.0.0-beta

"v2" is the internal milestone name for the Go rewrite. It is not the public version number, and it never will be.

Public reconFTW tags are already at v4.1. A v2.0.0-beta tag would sort below the current public release: every tool that orders versions — package managers, update checkers, humans skimming a releases page — would read it as a downgrade, and some would refuse to install it over an existing v4.x. So the rewrite's first public tag continues the public sequence instead of restarting it: v5.0.0-beta.1.

Expect v5.0.0-beta.2, -beta.3 and so on during the beta period, and v5.0.0 when it is no longer a beta.


You are not affected unless you opt in

releases/latest continues to resolve to bash v4.1.

That is not a promise we are keeping by hand — it is how GitHub works. GitHub never points releases/latest at a pre-release, and the release pipeline marks this build as a pre-release. The install command documented at the top of the README resolves through releases/latest, so it keeps serving the bash release for the entire beta period.

If you do nothing, nothing changes for you. To get the beta you have to name the tag explicitly — see the opt-in section in the README. The Go binary is reconftw and the bash entry point is reconftw.sh; they do not overwrite each other, so you can install the beta alongside your existing setup and fall back at any time by just running the other one.

Migration details, and every breaking change with a before/after example, are in MIGRATION.md — start at §0.


Where feedback goes

Use the v2 beta feedback issue template (.github/ISSUE_TEMPLATE/v2-beta-feedback.md). It applies the v2-beta and feedback labels automatically, which is what makes the triage below possible.

Bugs in the bash release still go to the normal Bug report template. The beta template is for the Go rewrite only.

Nothing in the template is required. A one-line "this broke" with the output of reconftw version is already useful.

Triage commitment (CUT-10): a triage pass over open v2-beta issues once a month, with the summary posted publicly as a repository issue tagged v2-beta — what came in, what was fixed, what was deferred and why. The first pass happens the month after the beta ships.

No triage has happened yet, because the beta has not shipped yet. This section describes a commitment, not a track record; judge it once there is something to judge.


Known issues already tracked

Please skip these — they are open on purpose, and reporting them adds noise rather than information:

  • Recon/<domain>/ is not created (CUT-08, above). Do tell us if it breaks a script — that specific report is wanted.
  • dnscewl and dnstake do not resolve on macOS in this beta.
  • dalfox may exit non-zero on --no-spinner if you have dalfox v3.x installed; the pinned argument vector targets v2.9.2.