uploading expectations: a clue-directed preview flow

huntress ctf 2026, day 22 · web exploitation · single, time trial, and mastery

tools used: live web target capture, saved capture inspection, and timed-attempt pipeline.

this challenge assigns a temporary upload application and asks for evidence from the rendered result. the route and preview path vary per target, so both must come from the current target's own responses.

the key insight is to read the clue document and preview guide before using the upload feature. a successful upload is only an intermediate result; the server-returned preview is where the useful evidence appears.

challenge shape

the single and time trial use one live target. mastery supplies three separate targets. each one publishes its own route data and returns its own preview path, so values from one target cannot be reused for another.

artifact anatomy

the landing page links to a route description and a preview guide. the clue document names the upload feature, while the guide describes the content the preview renderer expects. an upload response returns a same-origin preview path that can be read with the same target session.

extraction method

parse the target's current route map, send the documented upload structure to that route, and then request the returned preview path. do not keep a route or preview path between assignments.

clues = read_json(capture.get(observed_clue_path))
uploaded = capture.post_json(
    clues.upload_route,
    build_documented_upload(clues, guide),
)
preview_path = read_json(uploaded).preview_path
capture.get(preview_path)

validation and answer formatting

use the response from the preview path returned by that target, and match it to the same target's question before submitting. the proof is [redacted]; copy the response value as stored without changing its form.

failed approaches and pivots

a normal upload could return a preview without exposing the requested evidence. the guide explained the renderer's expected content, and the clue document supplied the current feature route. following both sources produced the proof in the preview.

automation across modes

the same short recipe handles each target: read its route map, submit one evidence-based upload, and follow the returned preview. the collector keeps each target's cookies and origin separate while running independent targets concurrently.

lessons

solve times

public scoreboard · day 22 finish: below top 10

  • 8:07.8single#52 of 161 solvers
  • 8:31.8time trial#46 of 157 solvers
  • 10:39.5mastery#46 of 155 solvers
  • #26team day finishof 145 teams that cleared it, in 10:39.5
  • #7single 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.

the timed rows use clean fresh assignments. the local records do not contain a single solve time for this target.

mode work to flag attempts wrong submits
single not recorded locally — —
time trial 5.47s 1 0
mastery 11.97s 1 0

the clock

timed window length is calculated from each saved assignment's started_at and deadline. used time is the successful attempt's measured start-to-flag time.

mode time available time used time left
time trial 123.2s 5.47s 117.7s
mastery 2m 6.8s 11.97s 114.8s

where we got stuck

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

{generated by a clanker}