fetch quest: requiring evidence before a target-side fetch
huntress ctf 2026, day 24 · web exploitation · all three modes
tools used: live web target capture and timed-attempt pipeline.
a fetch feature can reach a route that refuses direct requests from outside its container. the safe solve is to confirm both behaviors from the assigned target, then ask that feature for one clue-named route.
challenge shape
single assigned one web target. mastery combined several independent web families, including backup, idor, sql injection, and an internal-fetch target. the same recipe did not apply to every target in the bundle.
artifact anatomy
the clue response named a feature route, a secondary route, and the framework used by the target. the feature rejected a request missing an absolute url. the secondary route rejected direct access with an internal-only response. those two responses established the only fetch path used by the solver.
extraction method
request the two clue-provided routes in order. continue only when the feature returns absolute_url_required with http 400 and the secondary route returns internal_only with http 403. then pass the second route through the feature using the target container's loopback listener:
internal_url = "http://127.0.0.1:8080" + secondary_route
response = client.get(feature_route, params={"url": internal_url})
the http client encodes the complete url parameter. the solver extracts proof from the feature response; it does not infer success from status alone.
validation and answer formatting
both routes must be relative to and remain on the assigned origin, with bounded lengths. the fetch is skipped unless the two expected response signatures appear on the same target. the loopback listener and secondary path come from the challenge's own target and clue response.
in mastery, each target receives a separate session and clue pass. the ordered assignment list maps each recovered proof to its own answer id.
failed approaches and pivots
an earlier recipe sent a same-origin url without first confirming the feature's behavior. the day 24 evidence showed an absolute-url requirement and an internal-only route. we added both response checks before using the target-side fetch, and a negative fixture confirms that the fetch is skipped when the internal-only response is missing.
the old mastery capture saved three accepted proofs but lacked the fourth proof-bearing response. a local live retest was not performed against the completed target, so that capture cannot prove the corrected recipe.
automation across modes
the handler reads the current target's routes, checks the response gates, and uses the same bounded extraction in single and mixed mastery. it does not reuse host or path values between attempts. time trial uses the same recipe with the assigned target and its own deadline.
lessons
- require the target to demonstrate the condition that justifies a special request before sending it.
- keep an ssrf path narrow: one assigned origin, one clue-provided route, and one response-validated request through loopback.
- in a bundle, finish known targets independently while investigating the remaining ones.
solve times
public scoreboard · day 24 finish: top 3
- 1:54.0single#13 of 156 solvers
- 2:20.5time trial#11 of 153 solvers
- 3:10.8mastery#4 of 150 solvers
- #3team day finishof 139 teams that cleared it, in 3:47.1
- #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 table measures from the first local start to the recorded team completion. these are not successful local-attempt runtimes: no timed proof capture for this challenge reached a local flag.
| mode | work to team completion | attempts | wrong submits |
|---|---|---|---|
| single | 123.8s | 1 | 0 |
| time trial | not measured; no local start (team completion 20.0s after the prior mode) | 0 | 0 |
| mastery | 39.1s | 1 | 0 |
the clock
no successful local timed attempt was retained, so the winning-attempt clock was not recorded.
| mode | time available | time used | time left |
|---|---|---|---|
| time trial | not recorded | not recorded | not recorded |
| mastery | not recorded | not recorded | not recorded |
where we got stuck
- the first single investigation tried a fetch behavior without validating its preconditions. the response evidence later supplied the necessary gates.
- the local mastery capture had no proof for its fourth target. three other target proofs were accepted, but the saved bytes could not establish the remaining answer.
- the timed handoff retried without starting a target. no request-phase measurements were saved for that mode.
flags: single, time trial, and mastery [redacted].
{generated by a clanker}