back at you: the whole message reversed, as a twelfth line
huntress ctf 2026, day 12 · warmups · all three modes
tools used: decoding engine and timed-attempt pipeline.
the prompt: "maybe we started at the wrong end." the single and the time trial were one line of text written backwards; the mastery was a zip of 58 blobs, each with twelve numbered lines in twelve encodings, the twelfth line reversed.
the key insight: day 12 was not an encoding at all. days 10 and 11 had been base32 and base64url, and we had prepared the rest of the base family; the twelfth line was a whole-text transform instead. the decoder treats a reversal as one more reading of a line, ranked by how the whole message reads, so the day 11 code read it unchanged.
challenge shape
single 1 field one reversed line
time trial 1 field one reversed line
mastery 58 fields 58 blobs of 12 numbered lines, line k in day k's form
the field is message (setK.message in mastery): "the exact answer slug its final non-empty decoded line instructs you to submit", a word-word-word-NN slug.
artifact anatomy
the single read, from its last character to its first, as the same sentence every warmup carries: a congratulation, then the instruction and the slug. a reversed slug keeps its shape but not its order (NN-drow-drow-drow), so a slug pattern applied to the raw text finds nothing.
each mastery blob is the message split into twelve chunks:
01<TAB>chunk in base64 (day 1)
02<TAB>chunk in rot13 (day 2)
...
11<TAB>chunk in bare base64url (day 11)
12<TAB>chunk reversed (day 12)
the twelfth chunk holds the end of the message, so the end of the slug sits in the reversed line, digits first: a line such as NN-drow-d reads backwards into the tail of the slug that line 11's base64url began.
extraction method
for a single message the decoder peels: it tries every transform it knows, keeps readings that are printable, and stops at the first reading that holds a slug of dictionary words. reverse is one of those transforms:
def reverse(text: str) -> str:
return text[::-1]
for the mastery the decoder reads lines, not the whole blob. every line gets every one-step reading (plain included), and a beam joins the readings left to right, scoring the whole message by letter pairs across line boundaries. a reversed chunk read plain is a run of backwards letter pairs; read reversed it continues the eleventh line's text, so the beam keeps it.
validation and answer formatting
the answer is the slug as the decoded text states it. a slug made of dictionary words is the decoder's acceptance test; a slug of non-words comes back marked low confidence, with the per-line readings, instead of being submitted as if it were sure.
failed approaches and pivots
the day 12 preparation bet on the base family: other base64 and base58 alphabets, base92, base100, xxencode, ebcdic hex. none was asked. nothing had to change on the day; the prediction for day 13 covers both whole-text transforms (inverted case, rail fence, words or letters reversed) and codecs combined with a reversal, in the decoding engine.
automation across modes
the same call answered all three modes: the single and the time trial are one line, and the mastery's line decoding handled 58 sets in about 0.3 seconds.
lessons
- a trend in a sequence of encodings is a hypothesis, not a schedule. keep every reading cheap enough to try them all.
- a reversal is one more reading of a line, judged by how the whole message reads, not by the line alone.
solve times
public scoreboard · day 12 finish: 1st
- 0:02.2singlefirst blood, +0.41 s
- 0:03.6time trialfirst blood, +1.4 s
- 0:05.6masteryfirst blood, +2.2 s
- #1team day finishof 166 teams that cleared it, in 1:06.5
- #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 | 1.4s | 1 | 0 |
| time trial | 1.4s | 1 | 0 |
| mastery | 2.0s | 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 | 30.0s | 0.6s | 29.4s |
| mastery | 30.0s | 1.4s | 28.6s |
the game master's ranked pool was empty on day 12, so both timed modes ran on its regular timed pool, outside the hall-of-fame ranking. the clock itself grew that day: 30 seconds for each mode, against 6 to 12 seconds on days 3 to 11.
where we got stuck
- nowhere. the reversal was already a reading, and each mode took one attempt.
flags: single, time trial, and mastery [redacted].
{generated by a clanker}