double spent: racing the last credit
huntress ctf 2026, day 27 · web exploitation · all three modes
tools used: live web target capture, saved web inspection, and timed-attempt pipeline.
the target exposed a one-credit redemption flow. two requests had to pass its balance check before either completed the debit.
challenge shape
single and time trial assigned one target each. mastery assigned multiple independently captured targets, which could mix web families. each proof belonged to its own target.
artifact anatomy
the current target's clue identified a rails application, a redemption route, and a secondary status route. before redemption, the status response reported one remaining credit and no overspend. the target session cookie tied the requests to the same balance.
extraction method
read the status route first and require the one-credit precondition. release exactly two json posts to the clue-provided redemption route together, with the established session cookie on both requests. then read the status route again and collect a proof only if it reports that the single credit was spent twice.
assert status()["credits"] == 1
responses = race_two_posts(feature_route, cookies=session_cookies)
assert len(responses) == 2
proof = status().get("proof")
validation and answer formatting
the two posts and both status reads stay on the assigned origin. a successful post response alone does not establish the race effect; the final status must contain the proof. submitted results are [redacted].
failed approaches and pivots
the initial shared collector sent a serial post before checking the status. that consumed the sole credit and left no way to reproduce the race on the same target. a fresh variant confirmed that the status precondition must come first and the posts must share the session.
automation across modes
the collector recognizes the rails clue and one-credit status before taking the race path. each target has its own cookie jar and balance; mixed mastery targets are classified independently and each proof remains paired with its answer id.
lessons
- check a one-shot precondition before any state-changing request.
- concurrency and session continuity are both necessary to test this race.
solve times
public scoreboard · day 27 finish: below top 10
- 2:50.4single#14 of 152 solvers
- 9:46.6time trial#33 of 146 solvers
- 9:56.8mastery#22 of 143 solvers
- #38team day finishof 127 teams that cleared it, in 34:43.1
- #7single 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 recorded work began with the assigned target. external completions and missing local winning captures are marked as such.
| mode | work to flag | attempts | wrong submits |
|---|---|---|---|
| single | 2m51.8s | 1 | 0 recorded |
| time trial | 6m55.9s | 1 | 0 recorded |
| mastery | 10.0s | 1 | 0 recorded |
the clock
the winning assignments recorded start and deadline. time used is the assignment start to the recorded flag, rounded to a tenth of a second.
| mode | time available | time used | time left |
|---|---|---|---|
| time trial | 123.3s | 6.1s | 117.2s |
| mastery | 123.9s | 9.4s | 114.5s |
where we got stuck
- the serial request in the first collector pass spent the credit. the fix made the clue-driven race the first state-changing action, then checked the secondary status for proof.
flags: single, time trial, and mastery [redacted].
{generated by a clanker}