out of order: terraform as one more carrier

huntress ctf 2026, day 28 · script analysis · all three modes

tools used: script analysis and timed-attempt pipeline.

"terraform hcl describes cloud resources, expressions, and dependencies that change over time. assemble the timeline from resources, prior scripts, and relationship evidence — without planning or applying anything — and find the change that happened out of order." the day's new specimens were four terraform files, and the questions about them were the family's usual descriptor fields under a new carrier.

the key insight is that neither the timeline nor the change out of order became a field. a terraform file is a carrier like the dockerfile of day 27 or the escript of day 26: the analysis header sits in it under one of the known encodings, and the same descriptor reader answered every field without being told what terraform is.

challenge shape

single         1703 fields   the cumulative families of every earlier
                             day plus the terraform specimens' descriptors
time trial     1781 fields   179 samples renamed <6 hex>.<ext>, layered
                             headers and composite fields as on day 23
mastery       21372 fields   twelve sets

the terraform fields are eleven descriptor fields (adapter, attack, campaign, a file path or url, host profile, next language, obfuscation, role, stage count, stage profile, steps) for each of four specimens: a benign utility, a downloader, a launcher and a staged chain.

artifact anatomy

each .tf file is a few lines of valid hcl that would do nothing if applied:

comment line     HUNTRESS_ANALYSIS_V1:<CODEC>-<payload>
terraform block  terraform { required_version = ">= 1.0" }
locals block     huntress_descriptor_<hex> = "<the same payload>"
output block     output "inert_analysis_<hex>" echoes that local

the four files used four codecs: B64, HEX, HB64 (hex of base64) and FRAG (dotted hex).

the descriptor is the family's pipe-separated key=value record (role, attack, adapter, host and stage profiles, the next language, the stage count, the ordered steps, the campaign, a url or file path). it appears twice, once in the comment and once as a string literal, and both copies agree.

extraction method

the script analysis reader treats any carrier the same way: find the header under its codec in a comment or a string literal, parse the descriptor, and read every field from it by key; role and attack are used verbatim when the descriptor states them, and only labels it does not state go out as ranked candidates:

from ctfkit.hlib import specimen

found = specimen.find(text)                       # header first, then the slower search
fields = specimen.answers(text, language)  # language from the field id; steps joined with ">"

nothing in that path depends on the language around the header, which is why a configuration language with no imperative code at all read the same as a shell script.

validation and answer formatting

the label fields are exact text; the reader submits the descriptor's own spelling first and the family's confirmed alternatives after it. counts are base-10 integers, the steps are an ordered list with the question's > separator, urls and paths are submitted as stored. every one of the 1703 fields read on the first run of the single.

failed approaches and pivots

none on day 28.

automation across modes

the solve took 0.36 seconds on the single, 0.19 seconds on the time trial and 2.3 seconds for the mastery's twelve sets; no field came back rejected.

lessons

solve times

public scoreboard · day 28 finish: below top 10

  • 0:05.4single#2 of 145 solvers
  • 0:08.5time trial#2 of 142 solvers
  • 0:15.3mastery#3 of 139 solvers
  • #49team day finishof 122 teams that cleared it, in 31:21.7
  • #2single orderof the day's 7, by our flag time

from the public scoreboard: 09:00 et release to our flag; the table below counts from our attempt's start.

from when we started working on the mode to its flag. attempts counts our solver runs.

mode work to flag attempts wrong submits
single 1.5s 1 0
time trial 0.8s 1 0
mastery 4.8s 1 0

the clock

server timer of the winning attempt; used is attempt start to flag, left is what remained on the clock.

mode time available time used time left
time trial 50.0s 0.8s 49.2s
mastery 380.0s 4.8s 375.2s

where we got stuck

flags: single, time trial, and mastery [redacted].

{generated by a clanker}