the butler did it: a jenkins pipeline, two bash features and a classic asp host

huntress ctf 2026, day 20 · script analysis · all three modes

tools used: script-analysis toolkit and timed-attempt pipeline.

a jenkins groovy pipeline that, in the prompt's words, "splits its groovy stages between controller and agent", beside the day's two advanced bash features (coproc blocks and mapfile over a process substitution) and a classic asp page hosting vbscript whose inert handoff names the next language. 1329 fields in the single, every graded family of the category so far. 80 of them were new: four groovy polyglot specimens, the jenkins pipeline host and the classic asp host at eleven fields each, and the two bash specimens at seven.

the key insight: nothing new to parse. each specimen carries the HUNTRESS_ANALYSIS_V1 descriptor under one of the carriers and codecs the earlier days taught, and the day's new families were the predicted next rows.

challenge shape

single       1329 fields   every family (days 1-20), one specimen each
time trial   1329 fields   the same on fresh specimens
mastery      6645 fields   five sets

artifact anatomy

the bash specimens declare coproc NAME { ... } and mapfile -t arr < <(cmd); their descriptors name the feature, the syntax primitive and the analysis steps. their materials use two codecs the reader already had: reversed_base64| (the string reversed, then base64) and base64_literal|. the classic asp page (<% ... %> with Language=VBScript) holds its descriptor in a B64- comment header and again in a server-side string, and hands off to jscript. the jenkins host is the same carrier in groovy: an HB64- header, a final string descriptor_<hex> with the same payload, and an inert one-stage pipeline { agent none; ... } block; the controller and agent split the prompt describes lives only in the descriptor's steps and next language.

extraction method

the script toolkit finds the descriptor in any carrier (a comment, a string literal, an escaped block), decodes it by its named codec and reads the role, attack mapping, steps and relations from it; the host page's language and next language come from the same descriptor.

validation and answer formatting

the descriptor's own version tag and the field vocabulary the grader confirmed on earlier days bound every label: the role question's enum (benign_utility through advanced_syntax), attack ids such as T1105 for a downloader, T1059 for a staged chain that hands off and T1059.004 for the bash specimens, and obfuscation profiles such as base64 and hex_base64; a stage the descriptor cannot support is reported as ranked candidates rather than asserted.

failed approaches and pivots

none.

automation across modes

the same descriptor read answers all 1329 fields, and the five mastery sets are the same families on fresh specimens; the replay of the three saved runs with the table-completeness check on reports no missing row.

lessons

solve times

public scoreboard · day 20 finish: below top 10

  • 0:03.2single#2 of 163 solvers
  • 0:05.1time trial#2 of 159 solvers
  • 0:08.2mastery#2 of 159 solvers
  • #25team day finishof 150 teams that cleared it, in 12:50.9
  • #3single 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.

mode work to flag attempts wrong submits
single 1.4s 1 0
time trial 0.8s 1 0
mastery 1.6s 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 170.0s 1.6s 168.4s

where we got stuck

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

{generated by a clanker}