template temper: a clue-directed spring expression

huntress ctf 2026, day 23 · web exploitation · single, time trial, and mastery

tools used: live web target capture and saved web inspection.

the assigned site offered a rendering feature and asked for a proof from its own response. the route changed with the target, so the clue document was part of the input to the solution.

the key insight was to supply a template expression that names the rendering context's objective. a request to the feature route without that expression only confirmed that the route existed.

challenge shape

single and time trial each supplied one temporary target. mastery supplied a bundle whose targets could use different web families. every question named its target, and each target required its own clue read and cookie jar.

artifact anatomy

the first page linked to /api/clues. that json identified Spring Boot and supplied a feature_route. the route was generated for the current assignment; it was never copied from a previous target.

extraction method

read the clue, validate that the feature route stays on the assigned origin, then send the expression as the template query value. inspect the rendered feature response for the single proof.

clues = get_json("/api/clues")
if clues.get("framework") == "Spring Boot":
    response = get(clues["feature_route"], params={"template": "{{objective}}"})
    proof = read_proof(response)

validation and answer formatting

the proof is copied from the response exactly and kept with that target's answer id. the submitted result is [redacted]. a proof from another mastery target cannot fill this one even if both responses look similar.

failed approaches and pivots

the first generic capture read the clues and the feature route but omitted the expression, so it found no proof. the saved response map showed the route; the rendering behavior described by the target supplied the missing query.

automation across modes

the same target-local rule runs after each fresh clue read. a mixed mastery bundle classifies each target independently and maps its captured proof back to the question that named it.

lessons

solve times

public scoreboard · day 23 finish: 1st

  • 3:47.6single#19 of 159 solvers
  • 7:41.3time trial#23 of 154 solvers
  • 8:30.1mastery#17 of 152 solvers
  • #1team day finishof 140 teams that cleared it, in 8:58.5
  • #6single 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.

the preserved completion record gives assignment start to flag for the single and mastery. it does not preserve the winning time trial's start or wrong-submit history; those entries are left unknown.

mode work to flag attempts wrong submits
single 3m 44.6s from recorded start 1 recorded unknown
time trial unknown unknown unknown
mastery 40.1s from recorded start 1 recorded unknown

the clock

the challenge listed a nominal 120-second timer for ranked timed attempts. the completion record contains no deadline, so time left is an estimate for mastery and unavailable for the time trial.

mode time available time used time left
time trial 120s configured unknown unknown
mastery 120s configured 40.1s from start about 79.9s, inferred

where we got stuck

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

{generated by a clanker}