{ "agent_trace": { "codex_logs_available_local_only": true, "raw_logs_published": false, "summary_path": "agent-summary.json" }, "codex_version": "codex-cli 0.131.0", "collected_at": "2026-05-19T20:23:04.339411+00:00", "docker_cpus": 16, "docker_memory": "30g", "eval": { "error_code": null, "failed_tests": [ "eval.tests.test_error_handling.test_starting_password_with_dictionary_error", "eval.tests.test_error_handling.test_starting_password_too_long_error", "eval.tests.test_error_handling.test_starting_password_too_short_error", "eval.tests.test_error_handling.test_starting_password_out_of_charset_error", "eval.tests.test_starting_password.test_starting_password_outside_charset", "eval.tests.test_starting_password.test_starting_password_with_dictionary", "eval.tests.test_starting_password.test_starting_password_too_short", "eval.tests.test_starting_password.test_starting_password_too_long", "tests.test_bruteforce.test_starting_password_too_long", "tests.test_cli.test_inputfile_directory", "tests.test_cli.test_startingpassword_with_dictionary", "tests.test_cli.test_startingpassword_at_minlen_boundary", "tests.test_harvest.test_startingPassword_with_dictionary", "tests.test_harvest.test_startingPassword_length_exceeds_maxPasswordLen", "tests.test_performance.test_resume_from_invalid_starting_password_wrong_length", "tests.test_performance.test_resume_from_invalid_starting_password_wrong_charset", "tests.test_performance.test_starting_password_with_dictionary_error", "tests.test_performance.test_empty_charset_file_error", "tests.test_zip_files.test_starting_password_with_dictionary", "tests.test_bruteforce.test_starting_password_with_dictionary_error", "tests.test_cli.test_startingpassword_at_maxlen_boundary", "tests.test_harvest.test_startingPassword_out_of_charset", "tests.test_bruteforce.test_starting_password_out_of_charset", "tests.test_bruteforce.test_starting_password_too_short", "tests.test_cli.test_startingpassword_out_of_charset", "tests.test_harvest.test_startingPassword_length_below_minPasswordLen", "tests.test_cli.test_startingpassword_exceeds_maxlen", "eval.tests.test_argument_parsing.TestPasswordDictionaryArgument.test_dictionary_is_directory", "eval.tests.test_argument_parsing.TestInputFileArgument.test_input_file_is_directory", "eval.tests.test_argument_parsing.TestCharsetFileArgument.test_charset_file_is_directory", "eval.tests.test_argument_parsing.TestStartingPasswordArgument.test_starting_password_with_dictionary", "eval.tests.test_argument_parsing.TestStartingPasswordArgument.test_starting_password_out_of_charset", "eval.tests.test_argument_parsing.TestStartingPasswordArgument.test_starting_password_too_long", "eval.tests.test_argument_parsing.TestStartingPasswordArgument.test_starting_password_too_short", "eval.tests.test_argparse_validation.test_inputfile_accepts_space_and_equals_forms[args3]", "eval.tests.test_argparse_validation.test_integer_option_rejects_overflow", "eval.tests.test_argparse_validation.test_value_starting_with_dash_is_rejected_for_startingpassword", "eval.tests.test_argparse_validation.test_duplicate_charset_flag_is_rejected_by_parser", "tests.test_help_baseline.TestHelpBaseline.test_invalid_flag_format", "eval.tests.test_zip_password_finder.TestPasswordLength.test_min_password_length_constraint", "eval.tests.test_zip_password_finder.TestDictionaryAttack.test_dictionary_password_not_found" ], "failed_tests_omitted": 0, "public_eval_path": "eval.json", "summary_path": "eval-summary.json", "test_branch_errors": {}, "test_records": 792, "warnings": [] }, "host_machine": "x86_64", "host_system": "Linux", "inference_mode": "paper-prompt-nointernet", "instance_id": "agourlay__zip-password-finder.704700d", "metrics": { "almost_resolved": "False", "cached_input_tokens": "1794560", "calls": "48", "codex_version": "codex-cli 0.131.0", "created_at": "2026-05-18T23:20:45.360323+00:00", "docker_cpus": "16", "docker_memory": "30g", "error_code": "", "estimated_cost_usd": "", "evaluator_problem": "False", "host_machine": "x86_64", "host_system": "Linux", "inference_mode": "paper-prompt-nointernet", "input_tokens": "1862252", "instance_id": "agourlay__zip-password-finder.704700d", "long_context_calls": "0", "model": "gpt-5.5", "n_resolved_tests": "640", "n_system_errors": "0", "n_tests": "680", "n_warnings": "0", "output_tokens": "33470", "paper_compliant": "False", "pricing_source": "", "reasoning_effort": "xhigh", "reasoning_output_tokens": "12285", "resolved": "False", "run_name": "cpx62-paperprompt-xhigh-shard-1-of-5-20260518T232006Z-agourlay-zip-password-finder", "run_version": "20260518T232006Z", "score": "0.9411764705882353", "session_ended_at": "2026-05-18T23:29:21.330Z", "session_started_at": "2026-05-18T23:20:53.474Z", "test_branch_errors": "{}", "total_tokens": "1895722", "wall_clock_seconds": "507" }, "model": "gpt-5.5", "package": { "contents": [ ".", "./.git", "./.git/COMMIT_EDITMSG", "./.git/HEAD", "./.git/branches", "./.git/config", "./.git/description", "./.git/hooks", "./.git/hooks/applypatch-msg.sample", "./.git/hooks/commit-msg.sample", "./.git/hooks/fsmonitor-watchman.sample", "./.git/hooks/post-update.sample", "./.git/hooks/pre-applypatch.sample", "./.git/hooks/pre-commit.sample", "./.git/hooks/pre-merge-commit.sample", "./.git/hooks/pre-push.sample", "./.git/hooks/pre-rebase.sample", "./.git/hooks/pre-receive.sample", "./.git/hooks/prepare-commit-msg.sample", "./.git/hooks/push-to-checkout.sample", "./.git/hooks/update.sample", "./.git/index", "./.git/info", "./.git/info/exclude", "./.git/logs", "./.git/logs/HEAD", "./.git/logs/refs", "./.git/logs/refs/heads", "./.git/logs/refs/heads/master", "./.git/objects", "./.git/objects/01", "./.git/objects/01/166275df61b5e03047b79c246832ac2947a2c1", "./.git/objects/1b", "./.git/objects/1b/d0b5f58f26e50423808574bf4d5ad347d0f39d", "./.git/objects/33", "./.git/objects/33/ee0b4a1cf5db0ab2a11fdd3490741e9ce4271c", "./.git/objects/44", "./.git/objects/44/b1eb774551df7d9ac2b2b6e9972ed92a8b5e2d", "./.git/objects/59", "./.git/objects/59/b072e4138c9066cf8a45bdb59af361cd062ee2", "./.git/objects/8b", "./.git/objects/8b/a774a1f2a43093fd7ac3d5b8d5f19595f487d4", "./.git/objects/99", "./.git/objects/99/6db05d75d1e013a8aa94369383ef013b3db78c", "./.git/objects/a8", "./.git/objects/a8/b72b03f4b75746a9e38df9b9d2e29f8676bd19", "./.git/objects/ae", "./.git/objects/ae/0246ec8082f43fed2e41fac62ed134b02362cc", "./.git/objects/db", "./.git/objects/db/ef71ae253d9ab9f0c51dc376437fe3929488ad", "./.git/objects/e5", "./.git/objects/e5/48d85c45d4b1aabbb72842d6dec350b5397795", "./.git/objects/f2", "./.git/objects/f2/3fdf9843c25b09f511a6e07d61933b9cfd72d0", "./.git/objects/info", "./.git/objects/pack", "./.git/refs", "./.git/refs/heads", "./.git/refs/heads/master", "./.git/refs/tags", "./LICENSE", "./README.md", "./assets", "./assets/2.test.txt.zip", "./assets/2.test.txt_1.zip", "./assets/3.test.txt.zip", "./assets/3.test.txt_1.zip", "./assets/4.test.txt.zip", "./assets/password-123.zip", "./assets/password-ABC.zip", "./assets/password-a1.zip", "./assets/password-test.zip", "./compile.sh", "./executable", "./zip_password_finder.py" ], "contents_omitted": 0, "submission_available_local_only": true }, "paper_compliant": false, "paper_mode": true, "reasoning_effort": "xhigh", "run_name": "cpx62-paperprompt-xhigh-shard-1-of-5-20260518T232006Z-agourlay-zip-password-finder", "run_version": "20260518T232006Z", "usage_audit": { "available_local_artifact": true, "public_path": "usage-audit.json" } }