mixed signals: an erlang escript as one more carrier
huntress ctf 2026, day 26 · script analysis · all three modes
tools used: script analysis, decoding search engine and timed-attempt pipeline.
"an erlang escript passes messages between processes, and some of its indicators look suspiciously familiar. trace the message flow, and separate the true cross-language indicators from deterministic collision decoys planted to look like old friends." the day's carrier was an erlang escript, beside every earlier day's samples. the single added four polyglot-erlang-<role> samples; the time trial and the mastery had day 23's shape (random hex stems, layered analysis headers, composite answers) with escripts among the carriers.
the key insight: as on days 24 and 25, our descriptor finder keys on the analysis header's marker, not on the language around it. erlang comments start with %, and that did not matter: the marker search never parses comments at all.
challenge shape
single 1615 fields every earlier sample's descriptors plus four
polyglot-erlang-<role> samples, eleven fields
each
time trial 1677 fields samples renamed <6 hex>.<ext>; composite
fields (url host, path leaf, campaign token)
as on day 23
mastery 16770 fields the same over ten bundles
the four roles were benign-utility, downloader, launcher and staged-chain. each sample's fields are the descriptor's: role, attack technique, adapter, campaign, a url or file path, stage count, steps, and the host, obfuscation and stage profiles and the next language.
artifact anatomy
an escript sample is four lines of erlang source, a module that is never started:
% <marker>:<CODEC>-<payload> a comment: the analysis header
-module(plan_<hex>). a module attribute, a fresh id per file
-export([build_plan/0]). one exported function of arity zero
build_plan() -> "<payload>". which returns the same payload as a string
there is no main/1, which an escript needs to run, no spawn, no ! send and no receive. the "message flow" of the prompt is the story the descriptor tells (its ordered steps and the next language in the handoff), not erlang processes in the file. the "collision decoys" were story as well: no field asked to tell a decoy from a true indicator, and every indicator field read from the descriptor.
decoded, the payload is the descriptor the questions ask about, v=3| followed by |-separated key=value pairs: language, role, verdict, attack technique, adapter, host, obfuscation and stage profiles, the next language, the stage count, the ordered steps, the campaign id and the url and file path.
the single's four escript headers used four different codecs (B64, HEX, HB64 and FRAG), all of them codecs that earlier polyglot samples used. in the time trial the samples were renamed to six hex digits (<stem>.escript) and every header used day 23's layered codec, LYR-<layers>-<payload>.
extraction method
the descriptor finder searches every member for the marker in any comment or string syntax, reads the codec name after it, undoes that codec (or, for LYR, each named layer from the last to the first) and accepts the result only when it parses as a descriptor. the function's string literal carries the same payload as the header, without the marker and the codec name in front of it, a free cross-check.
from ctfkit.hlib import specimen
for member, text in members.items():
found = specimen.header_descriptor(text) # marker, codec name, payload; LYR layer by layer
if found is not None:
answers[member] = found.desc # the parsed key=value descriptor
every field is read from the descriptor, and a value the descriptor states is used verbatim. the time trial's composite fields came from day 23's composite reader: the campaign token, the url's host and the file path's last segment, each read by the reader of the standalone field with the same name.
validation and answer formatting
the decoded text must parse as a descriptor (|-separated key=value pairs) or the reading is refused, and a layer name the layered reader does not know raises a named error instead of being guessed around. the profiles and the next language are submitted as the descriptor states them; for the single's four escripts:
| sample | host profile | obfuscation | stage profile | next language |
|---|---|---|---|---|
benign-utility |
standalone |
hex_base64 |
three_stage_decode_write_launch |
powershell |
downloader |
standalone |
fragmented_hex |
cross_language_handoff |
jscript |
launcher |
standalone |
plain_hex |
single_descriptor |
batch |
staged-chain |
standalone |
base64 |
two_stage_download_execute |
vbscript |
the time trial and the mastery asked the same of their escripts except the host profile: four in the time trial (stage profile umber_ladder and next language awk for the first) and forty over the mastery's ten bundles (onyx_lattice and fish for the first).
failed approaches and pivots
none on day 26.
automation across modes
the same reader answered the single's 1615 fields, the time trial's 1677 and the mastery's 16770 across ten bundles; the mastery's solve took 2.42 seconds, and sending its answers took 1.59 more.
lessons
- a new comment character costs nothing when the finder looks for the marker rather than for comments; the language never has to be parsed.
- a prompt's story (processes, messages, decoys) can be richer than its questions; read the field list before modelling the story.
solve times
public scoreboard · day 26 finish: below top 10
- 0:03.9singlefirst blood, +0.87 s
- 0:06.8time trialfirst blood, +0.48 s
- 0:13.0mastery#2 of 145 solvers
- #32team day finishof 135 teams that cleared it, in 19:21.9
- #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 | 1.7s | 1 | 0 |
| time trial | 1.0s | 1 | 0 |
| mastery | 4.3s | 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 | 1.0s | 49.0s |
| mastery | 320.0s | 4.3s | 315.7s |
where we got stuck
- nowhere on day 26: the marker search found the escript headers on the first run, and the time trial's layered headers and composites were day 23's.
flags: single, time trial, and mastery [redacted].
{generated by a clanker}