hex appeal: solve each encoded line independently

huntress ctf 2026, day 3 · warmups · all three modes

tools used: decoding engine and timed-attempt pipeline.

day 3's codec was hexadecimal ("sixteen seems like a perfectly reasonable number of fingers"). the single and the time trial were the whole sentence as one run of lowercase hex digits, two per byte, no separators. mastery had thirteen sets, and each set's blob split the sentence across three numbered lines: base64, rot13 and hex, each line encoded on its own.

the key insight: the abstraction is no longer "decode this file" but "decode each record, then reconstruct the message".

artifact anatomy

each non-empty line has the form:

NN<TAB>encoded chunk

line numbers order the message; the codec can differ on every line. on day 3 the codec of line k was day k's: 01 base64, 02 rot13, 03 hex. the cut points fall mid-word, so the decoded chunks join with no separator, and the hex line carries the end of the sentence, slug included:

01	TmljZSB3b3JrIGRlY29kaW5nIHRoaXMgd2FybXVwISBUbyBz
02	byir, fhozvg guvf rknpg nafjre gb gu
03	652047616d65204d61737465723a20...

line 1 decodes to Nice work decoding this warmup! To s, line 2 to olve, submit this exact answer to th, line 3 to e game master: and the slug.

extraction method

records = sorted((int(number), chunk) for number, chunk in parsed_lines)
options = [candidate_decodings(chunk) for _, chunk in records]
for combination in best_first_product(options):
    joined = "".join(text for _, text in combination)
    if english_score(joined) and SLUG.search(joined):
        return joined

each chunk gets its supported single-step decodings; only printable results survive, a few candidates per line are kept, and joined combinations are tried best-english first.

validation and answer formatting

the answer is the last bounded match of [a-z]+-[a-z]+-[a-z]+-\d{2} in the reconstructed text (nimble-scooter-boulder-88 for the single, jovial-island-copper-93 for the time trial and clever-daisy-ember-99, lucky-button-hazel-43, gentle-canoe-nugget-50 and ten more for the mastery's thirteen sets), with the surrounding text reading as english.

failed approaches and pivots

three easy mistakes, each of which we either made or guarded against: lexicographic sorting puts line 10 before line 2; decoding the whole file mixes incompatible alphabets and keeps the line numbers; taking the first decoder that succeeds repeats day 2's rot13 ambiguity.

automation across modes

the solver itself became tiny: pick the non-archive blob, decode it as text, call the slug finder, and map the slug to every local field. mastery's thirteen sets ran through the same function, one slug per set.

lessons

solve times

public scoreboard · day 3 finish: 1st

  • 0:26.5single#3 of 336 solvers
  • 0:35.3time trial#3 of 309 solvers
  • 1:16.1masteryfirst blood, +19.9 s
  • #1team day finishof 199 teams that cleared it, in 14:26.6
  • #1single orderof the day's 5, 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 14s 1 0
time trial 5s 1 0
mastery 39s 2 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 6.0s 0.5s 5.5s
mastery 7.0s 1.9s 5.1s

where we got stuck

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

{generated by a clanker}