fail open friday: a policy decoder's documented failure path

huntress ctf 2026, day 28 · web exploitation · all three modes

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

the target guarded a feature behind a remote policy input. its own clue and format responses disclosed how to submit that input. the proof appeared when the documented prefix reached the decoder without a payload.

challenge shape

single and time trial each assigned one temporary target; mastery assigned five independently captured targets. each target had a proof field, and the mastery bundle could mix web families.

artifact anatomy

the clue document named a feature route and a policy-format route. the first feature request showed that a remote policy was required. the policy-format response gave a query parameter and an accepted prefix. those two fields supplied the request shape; the route alone did not.

extraction method

the solver follows the current target's format route and reads its query parameter and prefix. it then requests the feature route with the documented prefix and an empty payload. the response is inspected for one proof.

clues = read_clues(target)
policy = get(clues["policy_format_route"])
response = get(clues["feature_route"], params={
    policy["query_parameter"]: policy["prefix"],
})
proof = one_proof(response)

validation and answer formatting

the collector restricts every route to the assigned origin. it uses parameter names and prefix text from the current target rather than values copied from the single. a proof must be present in the response body; a success status is not enough. submitted results are [redacted].

failed approaches and pivots

the first revision corrected the format route but ignored the fields that route returned. another capture therefore found no proof. the second revision included both the disclosed parameter and prefix, and the single succeeded.

automation across modes

each target starts with its own clue and policy document. the capture chooses this flow only when those responses have the expected shape. proofs remain associated with their original target in a mixed mastery bundle.

lessons

solve times

public scoreboard · day 28 finish: below top 10

  • 1:15.2single#16 of 149 solvers
  • 1:27.1time trial#11 of 145 solvers
  • 1:50.3mastery#3 of 131 solvers
  • #49team day finishof 122 teams that cleared it, in 31:21.7
  • #3single 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.

work began when the release assigned the target. the timed rows measure from the preceding mode's flag.

mode work to flag attempts wrong submits
single 1m16.6s 1 0
time trial 16.4s 1 0
mastery 18.4s 1 0

the clock

the winning assignments recorded their start and deadline. time used is assignment start to flag.

mode time available time used time left
time trial 124.2s 15.7s 108.5s
mastery 131.5s 17.8s 113.7s

where we got stuck

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

{generated by a clanker}