pentest-public-51classes-20260611
OhanaSec/jc-pentest-harness · api-multi-model
Status
completed
Storage
neon_persisted
Spend
$2.84
Updated
Jun 11, 2026, 02:39 PM
Supervised Execution Boundary
A real run can create the scoped run record and prompt package, but Hunt, Validate, Gapfill, Trace, Report, and comparative scorecard completion require real stage artifacts. The UI does not mark those complete without evidence.
- 01 / setupconfigured
Operator setup and responsible-use gate.
- 02 / scope validationconfigured
Allowed hosts and target scope validated before launch.
- 03 / reconmanual required
Runner scaffolds the task queue.
- 04 / huntmanual required
Supervised model/client output is required.
- 05 / validatemanual required
Validate can confirm, downgrade, or reject only.
- 06 / gapfillmanual required
Refusals or uncovered cells require operator review.
- 07 / tracemanual required
Reachability analysis links findings to trust boundaries.
- 08 / reportmanual required
Reports are generated from real findings only.
- 09 / scorecardmanual required
Comparative scoring is a supervised rubric stage.
Scorecards
opus48artifact backed
83.4
11 confirmed / 11 findings
fable5artifact backed
77.0
8 confirmed / 9 findings
gpt55artifact backed
70.6
6 confirmed / 8 findings