eighty-five alive: ascii85 inside adobe's brackets
huntress ctf 2026, day 23 · warmups · all three modes
tools used: decoding search engine and timed-attempt pipeline.
"the alphabet got bigger. adobe might recognize the neighbors." day 22's newest line was base64 of a zlib stream; day 23's was ascii85, the 85-symbol encoding postscript and pdf use, wrapped in the <~ and ~> delimiters adobe's tools write around it. the single and the time trial were one such blob each; the mastery was a zip of 113 blobs of twenty-three numbered lines, one line per day of the event, whose newest line was the same framed ascii85.
the key insight: the delimiters are the neighbors the prompt means, and they name the variant. several base-85 alphabets exist; only adobe's ascii85 is written between <~ and ~>, so the framing picks the decoder before any search has to.
challenge shape
single 1 field the slug one blob's decoded text instructs
time trial 1 field the same, a fresh blob
mastery 113 fields one slug per blob, 113 blobs of 23 numbered lines
the field is message (setK.message in mastery).
artifact anatomy
the single's blob is one line of printable characters: <~, a run of characters between ! and u, and ~>. ascii85 maps every four bytes to five characters: the four bytes are one big-endian 32-bit number, written as five base-85 digits, each digit plus 33 (!). a group of four zero bytes may be written as the single character z, and a final group of one to three bytes is padded with zeros, encoded, and cut back to one character more than the bytes it carries. "the alphabet got bigger" is the step from base64's 64 symbols to these 85.
in the mastery each blob is NN<tab>chunk lines: one message split evenly over 23 lines, line k encoded by day k's encoding. line 23 is the framed ascii85; lines 1 to 22 are the earlier days' encodings, read by the schedule.
extraction method
the decoding engine already had an ascii85 reading, which turns adobe mode on when the text carries the opening delimiter (adobe mode strips <~ and ~> and refuses text outside them):
def a85(text: str) -> str:
return as_text(base64.a85decode(compact(text), adobe="<~" in text))
in the mastery, line k is read by day k's row of the schedule first, and only a line past the schedule is searched. the prompt gives the search no head start: it spells the number out ("eighty-five"), the hint table's words for the base-85 family are 85 and ascii85, and "adobe" is not a hint word (the table matches only octal in this prompt, through the "eight" inside "eighty"). the line fell to the next tier, every one-step reading in turn, where a85 was the only reading that ended the message in a slug. the schedule now names row 23 as a85.
validation and answer formatting
the other base-85 alphabets (z85, the rfc 1924 alphabet) share most of ascii85's characters, so a wrong alphabet still yields bytes. the message's own shape decides: the decoded text must read as english and end in a slug of the shape the question states ("three lowercase words separated by hyphens, followed by a hyphen and a two-digit number; submit the slug only"). the framed ascii85 reading is the only one that does.
failed approaches and pivots
none on day 23.
automation across modes
the same engine reads the single and time-trial blobs whole and every mastery blob line by line; the mastery's 113 blobs took about half a second.
lessons
- delimiters are evidence:
<~ ~>is adobe's ascii85,begin 644is uuencode,=ybeginis yenc. read the framing before trying alphabets. - a prompt's wordplay helps a person more than a word table: "eighty-five" and "adobe" point a reader at ascii85, while our hint words matched neither, and the message's shape still had to check out.
solve times
public scoreboard · day 23 finish: 1st
- 0:02.2single#2 of 174 solvers
- 0:04.4time trial#3 of 165 solvers
- 0:06.4mastery#2 of 162 solvers
- #1team day finishof 140 teams that cleared it, in 8:58.5
- #1single orderof the day's 8, 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 | 1.3s | 1 | 0 |
| time trial | 0.7s | 1 | 0 |
| mastery | 1.0s | 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 | 0.7s | 29.3s |
| mastery | 130.0s | 1.0s | 129.0s |
where we got stuck
- nowhere on day 23: the encoding and its framing were already readings of the engine.
flags: single, time trial, and mastery [redacted].
{generated by a clanker}