back up a minute: robots.txt to an indexed backup

huntress ctf 2026, day 19 web track · web exploitation · all three modes

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

the target was a small status site with no downloadable artifact. the task was to recover a value left in an exposed deployment backup.

the key insight was that neither the directory nor the backup filename needed guessing. the site disclosed the directory in robots.txt, and directory indexing disclosed the backup file inside it.

challenge shape

the single and time trial each assigned one temporary target. mastery assigned two fresh targets. the site varied the hidden directory and backup filename, so a reusable solution had to follow evidence instead of carrying a known path forward.

artifact anatomy

the home page looked like an ordinary deployment status page. its useful clue pointed toward crawler policy. /robots.txt contained a Disallow entry for an otherwise unlinked directory. requesting that directory returned an html index with a release-notes file and one json backup whose name and directory changed between targets.

the backup was valid json with ordinary deployment settings and a support field containing [redacted].

extraction method

the evidence chain was four same-origin requests:

GET /
GET /robots.txt
GET <disallowed directory>/
GET <linked backup file>

the solver parses only concrete Disallow paths, ignores /, reads links from the returned index, and follows files with a backup suffix:

directories = re.findall(r"(?im)^Disallow:\\s*(/[^\\s*?#]+)", robots)
for directory in directories[:3]:
    listing = get(directory)
    for link in Page(listing).links:
        if urlsplit(link).path.lower().endswith((".bak", ".backup", ".old")):
            get(urljoin(directory, link))

every resolved url must remain on the assigned origin. the collector records the responses once, and the answer step works from those saved bytes.

validation and answer formatting

the solution required exactly one requested value across the target's captured responses. no value was accepted from the directory name, filename or release notes. zero matches produced a diagnostic capture; multiple matches were an error rather than a reason to choose the first.

the submitted result in every mode was [redacted].

failed approaches and pivots

automation across modes

the same bounded recipe handled every target. mastery targets were captured in parallel with isolated sessions, then restored to target order before their questions were answered.

lessons

solve times

public scoreboard · day 19 finish: top 3

  • 1:37.1single#2 of 183 solvers
  • 2:58.7time trial#2 of 170 solvers
  • 17:04.2mastery#8 of 169 solvers
  • #3team day finishof 153 teams that cleared it, in 17:56.2
  • #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.

from when we started working on the mode to its flag.

mode work to flag attempts wrong submits
single 1m37s 1 0
time trial 13s 1 0
mastery 15s 1 0

the clock

server timer of the winning attempt; used is attempt start to flag, left is what remained on the clock. the accepted mastery run was untimed.

mode time available time used time left
time trial 120s 13s 107s

where we got stuck

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

{generated by a clanker}