reprosieve 0.1.0a3__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- reprosieve-0.1.0a3/.gitignore +17 -0
- reprosieve-0.1.0a3/CHANGELOG.md +66 -0
- reprosieve-0.1.0a3/LICENSE +202 -0
- reprosieve-0.1.0a3/PKG-INFO +214 -0
- reprosieve-0.1.0a3/README.md +179 -0
- reprosieve-0.1.0a3/SECURITY.md +14 -0
- reprosieve-0.1.0a3/pyproject.toml +83 -0
- reprosieve-0.1.0a3/schemas/capsule-v1.schema.json +62 -0
- reprosieve-0.1.0a3/schemas/case-study-v1.schema.json +306 -0
- reprosieve-0.1.0a3/schemas/materialization-v1.schema.json +21 -0
- reprosieve-0.1.0a3/schemas/predicate-report-v1.schema.json +35 -0
- reprosieve-0.1.0a3/schemas/reduction-report-v1.schema.json +29 -0
- reprosieve-0.1.0a3/src/reprosieve/__init__.py +3 -0
- reprosieve-0.1.0a3/src/reprosieve/_capture_bootstrap.py +67 -0
- reprosieve-0.1.0a3/src/reprosieve/adapters/__init__.py +0 -0
- reprosieve-0.1.0a3/src/reprosieve/adapters/openai_agents.py +598 -0
- reprosieve-0.1.0a3/src/reprosieve/adapters/openai_agents_replay.py +913 -0
- reprosieve-0.1.0a3/src/reprosieve/capsule.py +444 -0
- reprosieve-0.1.0a3/src/reprosieve/cli.py +414 -0
- reprosieve-0.1.0a3/src/reprosieve/ddmin.py +78 -0
- reprosieve-0.1.0a3/src/reprosieve/export.py +703 -0
- reprosieve-0.1.0a3/src/reprosieve/fixtures.py +121 -0
- reprosieve-0.1.0a3/src/reprosieve/hierarchy.py +664 -0
- reprosieve-0.1.0a3/src/reprosieve/predicate.py +598 -0
- reprosieve-0.1.0a3/src/reprosieve/py.typed +1 -0
- reprosieve-0.1.0a3/src/reprosieve/redact.py +244 -0
- reprosieve-0.1.0a3/src/reprosieve/replay.py +87 -0
- reprosieve-0.1.0a3/src/reprosieve/safeio.py +155 -0
- reprosieve-0.1.0a3/src/reprosieve/schema.py +253 -0
- reprosieve-0.1.0a3/src/reprosieve/verify.py +162 -0
|
@@ -0,0 +1,66 @@
|
|
|
1
|
+
# Changelog
|
|
2
|
+
|
|
3
|
+
All notable changes will be recorded here.
|
|
4
|
+
|
|
5
|
+
## 0.1.0a3 - release candidate
|
|
6
|
+
|
|
7
|
+
- Made the tag-bound package proof pass an absolute output path so its
|
|
8
|
+
clean-checkout install smoke tests can resolve the candidate artifacts.
|
|
9
|
+
- Supersedes the unpublished `v0.1.0a2` candidate, whose release workflow
|
|
10
|
+
stopped before any registry upload.
|
|
11
|
+
|
|
12
|
+
## 0.1.0a2 - unpublished candidate
|
|
13
|
+
|
|
14
|
+
- Kept release evidence downloads outside the source checkout so the
|
|
15
|
+
reproducible package proof runs from a clean tree.
|
|
16
|
+
- Supersedes the unpublished `v0.1.0a1` candidate, whose release workflow
|
|
17
|
+
stopped before any registry upload.
|
|
18
|
+
|
|
19
|
+
## 0.1.0a1 - unpublished candidate
|
|
20
|
+
|
|
21
|
+
- Renamed the unpublished public product, distribution, module, executable,
|
|
22
|
+
capsule suffix, and repository target to ReproSieve / `reprosieve` after the
|
|
23
|
+
final collision check found an active project using RunSieve / `runsieve`.
|
|
24
|
+
Immutable contract-v2 records and `RUNSIEVE_*` evidence protocol variables
|
|
25
|
+
retain their historical identifiers.
|
|
26
|
+
- Added public OpenAI Agents SDK trace capture for version 0.18.3.
|
|
27
|
+
- Added bounded before-disk redaction and deterministic validated capsules.
|
|
28
|
+
- Added constrained recorded-output materialization and strict tri-state/K-of-N predicate trials.
|
|
29
|
+
- Added dependency-aware hierarchical reduction and independent 1-minimality verification.
|
|
30
|
+
- Added hash-addressed artifacts, standalone issue exports, the 247-event demo, and security gates.
|
|
31
|
+
- Delayed capture publication until processor shutdown so multi-trace runs fail closed.
|
|
32
|
+
- Preserved declared environment entries in standalone reproductions.
|
|
33
|
+
- Hardened offline network audit denial and bounded user regex validation.
|
|
34
|
+
- Replaced self-attested generic gates with gate-specific measured command and artifact validation.
|
|
35
|
+
- Added commit-bound Linux/macOS clean-room proof bundles for RS-G10.
|
|
36
|
+
- Replaced the experimental adapter claim with explicit `materialize`,
|
|
37
|
+
`reproduce-predicate`, and `reduce` commands. Application replay is deferred;
|
|
38
|
+
pre-0.1 `replay` and `minimize` remain warning aliases.
|
|
39
|
+
- Allowed only operating-system temporary-directory symlink prefixes and added a reviewed secret-scan baseline.
|
|
40
|
+
- Added a corrected unrelated contract-v2 lineage whose immutable terminology
|
|
41
|
+
matches recorded-output materialization and offline predicate reproduction.
|
|
42
|
+
- Added an experimental, public-API OpenAI Agents application-replay adapter
|
|
43
|
+
with ordered exact substitution, measured canaries, divergence failures, and
|
|
44
|
+
independently verified synthetic evidence. It remains outside the 0.1
|
|
45
|
+
promise and is not 0.5-ready without permissioned real-case evidence.
|
|
46
|
+
- Added a strict permissioned case-study package schema and structural verifier
|
|
47
|
+
while recording all three real-case categories as external blockers.
|
|
48
|
+
- Closed capsule and capture time-of-check/time-of-use windows with bounded
|
|
49
|
+
single-open reads; rejected portable case, Unicode, trailing-dot, device-name,
|
|
50
|
+
and archive-member aliases.
|
|
51
|
+
- Made reduction reports byte-deterministic, corrected raw stdout/stderr
|
|
52
|
+
digests, hardened Windows process denial and tree termination, and validated
|
|
53
|
+
standalone reproduction resource policies before execution.
|
|
54
|
+
- Added replay-session tool ownership, expanded independent divergence evidence,
|
|
55
|
+
and added an exact final-unit minimality oracle.
|
|
56
|
+
- Added artifact-bound final evidence, wheel/sdist semantic and installed-flow
|
|
57
|
+
parity, deterministic SPDX and checksum artifacts, exact annotated-tag
|
|
58
|
+
preflight, and a separate OIDC-only PyPI publication job.
|
|
59
|
+
- Completed the non-destructive contract-v2 canonical migration, required CI
|
|
60
|
+
and tag protections, independent AI technical review, and public-default
|
|
61
|
+
fresh-clone proof.
|
|
62
|
+
- Required explicit authorization before exported capsules execute embedded
|
|
63
|
+
Python; denied `os.exec` as defense in depth.
|
|
64
|
+
- Made export, reduction, and verification consume one immutable validated
|
|
65
|
+
capsule snapshot; aligned standalone and package validation; rejected
|
|
66
|
+
undefined archive members and environment-selected temporary symlink roots.
|
|
@@ -0,0 +1,202 @@
|
|
|
1
|
+
|
|
2
|
+
Apache License
|
|
3
|
+
Version 2.0, January 2004
|
|
4
|
+
http://www.apache.org/licenses/
|
|
5
|
+
|
|
6
|
+
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
|
|
7
|
+
|
|
8
|
+
1. Definitions.
|
|
9
|
+
|
|
10
|
+
"License" shall mean the terms and conditions for use, reproduction,
|
|
11
|
+
and distribution as defined by Sections 1 through 9 of this document.
|
|
12
|
+
|
|
13
|
+
"Licensor" shall mean the copyright owner or entity authorized by
|
|
14
|
+
the copyright owner that is granting the License.
|
|
15
|
+
|
|
16
|
+
"Legal Entity" shall mean the union of the acting entity and all
|
|
17
|
+
other entities that control, are controlled by, or are under common
|
|
18
|
+
control with that entity. For the purposes of this definition,
|
|
19
|
+
"control" means (i) the power, direct or indirect, to cause the
|
|
20
|
+
direction or management of such entity, whether by contract or
|
|
21
|
+
otherwise, or (ii) ownership of fifty percent (50%) or more of the
|
|
22
|
+
outstanding shares, or (iii) beneficial ownership of such entity.
|
|
23
|
+
|
|
24
|
+
"You" (or "Your") shall mean an individual or Legal Entity
|
|
25
|
+
exercising permissions granted by this License.
|
|
26
|
+
|
|
27
|
+
"Source" form shall mean the preferred form for making modifications,
|
|
28
|
+
including but not limited to software source code, documentation
|
|
29
|
+
source, and configuration files.
|
|
30
|
+
|
|
31
|
+
"Object" form shall mean any form resulting from mechanical
|
|
32
|
+
transformation or translation of a Source form, including but
|
|
33
|
+
not limited to compiled object code, generated documentation,
|
|
34
|
+
and conversions to other media types.
|
|
35
|
+
|
|
36
|
+
"Work" shall mean the work of authorship, whether in Source or
|
|
37
|
+
Object form, made available under the License, as indicated by a
|
|
38
|
+
copyright notice that is included in or attached to the work
|
|
39
|
+
(an example is provided in the Appendix below).
|
|
40
|
+
|
|
41
|
+
"Derivative Works" shall mean any work, whether in Source or Object
|
|
42
|
+
form, that is based on (or derived from) the Work and for which the
|
|
43
|
+
editorial revisions, annotations, elaborations, or other modifications
|
|
44
|
+
represent, as a whole, an original work of authorship. For the purposes
|
|
45
|
+
of this License, Derivative Works shall not include works that remain
|
|
46
|
+
separable from, or merely link (or bind by name) to the interfaces of,
|
|
47
|
+
the Work and Derivative Works thereof.
|
|
48
|
+
|
|
49
|
+
"Contribution" shall mean any work of authorship, including
|
|
50
|
+
the original version of the Work and any modifications or additions
|
|
51
|
+
to that Work or Derivative Works thereof, that is intentionally
|
|
52
|
+
submitted to Licensor for inclusion in the Work by the copyright owner
|
|
53
|
+
or by an individual or Legal Entity authorized to submit on behalf of
|
|
54
|
+
the copyright owner. For the purposes of this definition, "submitted"
|
|
55
|
+
means any form of electronic, verbal, or written communication sent
|
|
56
|
+
to the Licensor or its representatives, including but not limited to
|
|
57
|
+
communication on electronic mailing lists, source code control systems,
|
|
58
|
+
and issue tracking systems that are managed by, or on behalf of, the
|
|
59
|
+
Licensor for the purpose of discussing and improving the Work, but
|
|
60
|
+
excluding communication that is conspicuously marked or otherwise
|
|
61
|
+
designated in writing by the copyright owner as "Not a Contribution."
|
|
62
|
+
|
|
63
|
+
"Contributor" shall mean Licensor and any individual or Legal Entity
|
|
64
|
+
on behalf of whom a Contribution has been received by Licensor and
|
|
65
|
+
subsequently incorporated within the Work.
|
|
66
|
+
|
|
67
|
+
2. Grant of Copyright License. Subject to the terms and conditions of
|
|
68
|
+
this License, each Contributor hereby grants to You a perpetual,
|
|
69
|
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
|
70
|
+
copyright license to reproduce, prepare Derivative Works of,
|
|
71
|
+
publicly display, publicly perform, sublicense, and distribute the
|
|
72
|
+
Work and such Derivative Works in Source or Object form.
|
|
73
|
+
|
|
74
|
+
3. Grant of Patent License. Subject to the terms and conditions of
|
|
75
|
+
this License, each Contributor hereby grants to You a perpetual,
|
|
76
|
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
|
77
|
+
(except as stated in this section) patent license to make, have made,
|
|
78
|
+
use, offer to sell, sell, import, and otherwise transfer the Work,
|
|
79
|
+
where such license applies only to those patent claims licensable
|
|
80
|
+
by such Contributor that are necessarily infringed by their
|
|
81
|
+
Contribution(s) alone or by combination of their Contribution(s)
|
|
82
|
+
with the Work to which such Contribution(s) was submitted. If You
|
|
83
|
+
institute patent litigation against any entity (including a
|
|
84
|
+
cross-claim or counterclaim in a lawsuit) alleging that the Work
|
|
85
|
+
or a Contribution incorporated within the Work constitutes direct
|
|
86
|
+
or contributory patent infringement, then any patent licenses
|
|
87
|
+
granted to You under this License for that Work shall terminate
|
|
88
|
+
as of the date such litigation is filed.
|
|
89
|
+
|
|
90
|
+
4. Redistribution. You may reproduce and distribute copies of the
|
|
91
|
+
Work or Derivative Works thereof in any medium, with or without
|
|
92
|
+
modifications, and in Source or Object form, provided that You
|
|
93
|
+
meet the following conditions:
|
|
94
|
+
|
|
95
|
+
(a) You must give any other recipients of the Work or
|
|
96
|
+
Derivative Works a copy of this License; and
|
|
97
|
+
|
|
98
|
+
(b) You must cause any modified files to carry prominent notices
|
|
99
|
+
stating that You changed the files; and
|
|
100
|
+
|
|
101
|
+
(c) You must retain, in the Source form of any Derivative Works
|
|
102
|
+
that You distribute, all copyright, patent, trademark, and
|
|
103
|
+
attribution notices from the Source form of the Work,
|
|
104
|
+
excluding those notices that do not pertain to any part of
|
|
105
|
+
the Derivative Works; and
|
|
106
|
+
|
|
107
|
+
(d) If the Work includes a "NOTICE" text file as part of its
|
|
108
|
+
distribution, then any Derivative Works that You distribute must
|
|
109
|
+
include a readable copy of the attribution notices contained
|
|
110
|
+
within such NOTICE file, excluding those notices that do not
|
|
111
|
+
pertain to any part of the Derivative Works, in at least one
|
|
112
|
+
of the following places: within a NOTICE text file distributed
|
|
113
|
+
as part of the Derivative Works; within the Source form or
|
|
114
|
+
documentation, if provided along with the Derivative Works; or,
|
|
115
|
+
within a display generated by the Derivative Works, if and
|
|
116
|
+
wherever such third-party notices normally appear. The contents
|
|
117
|
+
of the NOTICE file are for informational purposes only and
|
|
118
|
+
do not modify the License. You may add Your own attribution
|
|
119
|
+
notices within Derivative Works that You distribute, alongside
|
|
120
|
+
or as an addendum to the NOTICE text from the Work, provided
|
|
121
|
+
that such additional attribution notices cannot be construed
|
|
122
|
+
as modifying the License.
|
|
123
|
+
|
|
124
|
+
You may add Your own copyright statement to Your modifications and
|
|
125
|
+
may provide additional or different license terms and conditions
|
|
126
|
+
for use, reproduction, or distribution of Your modifications, or
|
|
127
|
+
for any such Derivative Works as a whole, provided Your use,
|
|
128
|
+
reproduction, and distribution of the Work otherwise complies with
|
|
129
|
+
the conditions stated in this License.
|
|
130
|
+
|
|
131
|
+
5. Submission of Contributions. Unless You explicitly state otherwise,
|
|
132
|
+
any Contribution intentionally submitted for inclusion in the Work
|
|
133
|
+
by You to the Licensor shall be under the terms and conditions of
|
|
134
|
+
this License, without any additional terms or conditions.
|
|
135
|
+
Notwithstanding the above, nothing herein shall supersede or modify
|
|
136
|
+
the terms of any separate license agreement you may have executed
|
|
137
|
+
with Licensor regarding such Contributions.
|
|
138
|
+
|
|
139
|
+
6. Trademarks. This License does not grant permission to use the trade
|
|
140
|
+
names, trademarks, service marks, or product names of the Licensor,
|
|
141
|
+
except as required for reasonable and customary use in describing the
|
|
142
|
+
origin of the Work and reproducing the content of the NOTICE file.
|
|
143
|
+
|
|
144
|
+
7. Disclaimer of Warranty. Unless required by applicable law or
|
|
145
|
+
agreed to in writing, Licensor provides the Work (and each
|
|
146
|
+
Contributor provides its Contributions) on an "AS IS" BASIS,
|
|
147
|
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
|
|
148
|
+
implied, including, without limitation, any warranties or conditions
|
|
149
|
+
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
|
|
150
|
+
PARTICULAR PURPOSE. You are solely responsible for determining the
|
|
151
|
+
appropriateness of using or redistributing the Work and assume any
|
|
152
|
+
risks associated with Your exercise of permissions under this License.
|
|
153
|
+
|
|
154
|
+
8. Limitation of Liability. In no event and under no legal theory,
|
|
155
|
+
whether in tort (including negligence), contract, or otherwise,
|
|
156
|
+
unless required by applicable law (such as deliberate and grossly
|
|
157
|
+
negligent acts) or agreed to in writing, shall any Contributor be
|
|
158
|
+
liable to You for damages, including any direct, indirect, special,
|
|
159
|
+
incidental, or consequential damages of any character arising as a
|
|
160
|
+
result of this License or out of the use or inability to use the
|
|
161
|
+
Work (including but not limited to damages for loss of goodwill,
|
|
162
|
+
work stoppage, computer failure or malfunction, or any and all
|
|
163
|
+
other commercial damages or losses), even if such Contributor
|
|
164
|
+
has been advised of the possibility of such damages.
|
|
165
|
+
|
|
166
|
+
9. Accepting Warranty or Additional Liability. While redistributing
|
|
167
|
+
the Work or Derivative Works thereof, You may choose to offer,
|
|
168
|
+
and charge a fee for, acceptance of support, warranty, indemnity,
|
|
169
|
+
or other liability obligations and/or rights consistent with this
|
|
170
|
+
License. However, in accepting such obligations, You may act only
|
|
171
|
+
on Your own behalf and on Your sole responsibility, not on behalf
|
|
172
|
+
of any other Contributor, and only if You agree to indemnify,
|
|
173
|
+
defend, and hold each Contributor harmless for any liability
|
|
174
|
+
incurred by, or claims asserted against, such Contributor by reason
|
|
175
|
+
of your accepting any such warranty or additional liability.
|
|
176
|
+
|
|
177
|
+
END OF TERMS AND CONDITIONS
|
|
178
|
+
|
|
179
|
+
APPENDIX: How to apply the Apache License to your work.
|
|
180
|
+
|
|
181
|
+
To apply the Apache License to your work, attach the following
|
|
182
|
+
boilerplate notice, with the fields enclosed by brackets "[]"
|
|
183
|
+
replaced with your own identifying information. (Don't include
|
|
184
|
+
the brackets!) The text should be enclosed in the appropriate
|
|
185
|
+
comment syntax for the file format. We also recommend that a
|
|
186
|
+
file or class name and description of purpose be included on the
|
|
187
|
+
same "printed page" as the copyright notice for easier
|
|
188
|
+
identification within third-party archives.
|
|
189
|
+
|
|
190
|
+
Copyright [yyyy] [name of copyright owner]
|
|
191
|
+
|
|
192
|
+
Licensed under the Apache License, Version 2.0 (the "License");
|
|
193
|
+
you may not use this file except in compliance with the License.
|
|
194
|
+
You may obtain a copy of the License at
|
|
195
|
+
|
|
196
|
+
http://www.apache.org/licenses/LICENSE-2.0
|
|
197
|
+
|
|
198
|
+
Unless required by applicable law or agreed to in writing, software
|
|
199
|
+
distributed under the License is distributed on an "AS IS" BASIS,
|
|
200
|
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
201
|
+
See the License for the specific language governing permissions and
|
|
202
|
+
limitations under the License.
|
|
@@ -0,0 +1,214 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: reprosieve
|
|
3
|
+
Version: 0.1.0a3
|
|
4
|
+
Summary: Reduce failed agent traces into redacted offline predicate reproductions
|
|
5
|
+
Project-URL: Homepage, https://github.com/DelshadH/reprosieve
|
|
6
|
+
Project-URL: Issues, https://github.com/DelshadH/reprosieve/issues
|
|
7
|
+
Project-URL: Source, https://github.com/DelshadH/reprosieve
|
|
8
|
+
Author: ReproSieve contributors
|
|
9
|
+
License-Expression: Apache-2.0
|
|
10
|
+
License-File: LICENSE
|
|
11
|
+
Classifier: Development Status :: 3 - Alpha
|
|
12
|
+
Classifier: Environment :: Console
|
|
13
|
+
Classifier: License :: OSI Approved :: Apache Software License
|
|
14
|
+
Classifier: Programming Language :: Python :: 3 :: Only
|
|
15
|
+
Classifier: Programming Language :: Python :: 3.11
|
|
16
|
+
Classifier: Programming Language :: Python :: 3.12
|
|
17
|
+
Classifier: Programming Language :: Python :: 3.13
|
|
18
|
+
Classifier: Topic :: Software Development :: Testing
|
|
19
|
+
Requires-Python: <3.14,>=3.11
|
|
20
|
+
Provides-Extra: dev
|
|
21
|
+
Requires-Dist: bandit==1.9.4; extra == 'dev'
|
|
22
|
+
Requires-Dist: build==1.5.0; extra == 'dev'
|
|
23
|
+
Requires-Dist: detect-secrets==1.5.0; extra == 'dev'
|
|
24
|
+
Requires-Dist: hypothesis==6.161.2; extra == 'dev'
|
|
25
|
+
Requires-Dist: mypy==2.3.0; extra == 'dev'
|
|
26
|
+
Requires-Dist: openai-agents==0.18.3; extra == 'dev'
|
|
27
|
+
Requires-Dist: pip-audit==2.10.1; extra == 'dev'
|
|
28
|
+
Requires-Dist: pip-licenses==5.5.5; extra == 'dev'
|
|
29
|
+
Requires-Dist: pytest-cov==7.1.0; extra == 'dev'
|
|
30
|
+
Requires-Dist: pytest==9.1.1; extra == 'dev'
|
|
31
|
+
Requires-Dist: ruff==0.16.0; extra == 'dev'
|
|
32
|
+
Provides-Extra: openai
|
|
33
|
+
Requires-Dist: openai-agents<0.19,>=0.18.3; extra == 'openai'
|
|
34
|
+
Description-Content-Type: text/markdown
|
|
35
|
+
|
|
36
|
+
# ReproSieve
|
|
37
|
+
|
|
38
|
+
**Turn a large failed agent run into a small, redacted, offline reproduction.**
|
|
39
|
+
|
|
40
|
+
ReproSieve captures one supported OpenAI Agents SDK trace, removes secrets before
|
|
41
|
+
persistence, replaces model and tool calls with recorded outputs, and reduces the
|
|
42
|
+
trace against an executable failure predicate. The result is a deterministic,
|
|
43
|
+
hash-addressed capsule with an independent 1-minimality proof.
|
|
44
|
+
|
|
45
|
+
> **Status:** honest pre-0.1 seed. The end-to-end path works and has synthetic
|
|
46
|
+
> security fixtures, but it has not been proven safe for real credentials,
|
|
47
|
+
> private source, or personal data. Use synthetic or disposable inputs.
|
|
48
|
+
|
|
49
|
+
## Supported path
|
|
50
|
+
|
|
51
|
+
- Python 3.11, 3.12, and 3.13.
|
|
52
|
+
- `openai-agents>=0.18.3,<0.19`; CI tests 0.18.3.
|
|
53
|
+
- One completed SDK trace captured through the public `TracingProcessor` API.
|
|
54
|
+
- Embedded Python predicates. ReproSieve invokes them with a direct argument vector,
|
|
55
|
+
a clean temporary directory, provider keys removed, bounded time/output/process
|
|
56
|
+
resources, and Python audit hooks that deny outbound network, child processes,
|
|
57
|
+
native loading, and host-file access. These controls are defense in depth, not
|
|
58
|
+
an OS sandbox; embedded predicates are arbitrary code.
|
|
59
|
+
- Deterministic recorded-output materialization and offline predicate
|
|
60
|
+
reproduction. These paths reconstruct retained values and execute only the
|
|
61
|
+
declared predicate; they do not rerun application or orchestration code.
|
|
62
|
+
|
|
63
|
+
## Support matrix
|
|
64
|
+
|
|
65
|
+
| Surface | Supported |
|
|
66
|
+
|---|---|
|
|
67
|
+
| Python | 3.11, 3.12, 3.13 |
|
|
68
|
+
| Capture SDK | `openai-agents>=0.18.3,<0.19` |
|
|
69
|
+
| Standalone reproduction | Linux and macOS |
|
|
70
|
+
| Input safety claim | Synthetic or disposable data only |
|
|
71
|
+
|
|
72
|
+
Install the core:
|
|
73
|
+
|
|
74
|
+
```bash
|
|
75
|
+
python -m pip install reprosieve
|
|
76
|
+
```
|
|
77
|
+
|
|
78
|
+
Install capture support:
|
|
79
|
+
|
|
80
|
+
```bash
|
|
81
|
+
python -m pip install "reprosieve[openai]"
|
|
82
|
+
```
|
|
83
|
+
|
|
84
|
+
## Workflow
|
|
85
|
+
|
|
86
|
+
The predicate script is arbitrary capsule-provided Python and is not sandboxed.
|
|
87
|
+
It must be included in the capsule, and every command that can execute or export
|
|
88
|
+
it requires `--trust-embedded-predicate`. `--predicate` must be the final option
|
|
89
|
+
because everything after it is an argument vector.
|
|
90
|
+
|
|
91
|
+
```bash
|
|
92
|
+
reprosieve capture \
|
|
93
|
+
--output failed.reprosieve \
|
|
94
|
+
--workspace-root . \
|
|
95
|
+
--include verify_failure.py \
|
|
96
|
+
-- python app.py
|
|
97
|
+
|
|
98
|
+
reprosieve reduce failed.reprosieve \
|
|
99
|
+
--output-dir reduced \
|
|
100
|
+
--trust-embedded-predicate \
|
|
101
|
+
--predicate python verify_failure.py
|
|
102
|
+
|
|
103
|
+
reprosieve materialize reduced/<sha256>.reprosieve \
|
|
104
|
+
--output materialized.json
|
|
105
|
+
|
|
106
|
+
reprosieve reproduce-predicate reduced/<sha256>.reprosieve \
|
|
107
|
+
--trust-embedded-predicate \
|
|
108
|
+
--predicate python verify_failure.py
|
|
109
|
+
|
|
110
|
+
reprosieve verify-minimal reduced/<sha256>.reprosieve \
|
|
111
|
+
--trust-embedded-predicate \
|
|
112
|
+
--predicate python verify_failure.py
|
|
113
|
+
|
|
114
|
+
reprosieve export reduced/<sha256>.reprosieve \
|
|
115
|
+
--trust-embedded-predicate \
|
|
116
|
+
--output issue-repro
|
|
117
|
+
|
|
118
|
+
cd issue-repro
|
|
119
|
+
python reproduce.py --trust-embedded-predicate
|
|
120
|
+
```
|
|
121
|
+
|
|
122
|
+
Exported reproductions refuse to execute without that explicit flag. Use it only
|
|
123
|
+
after inspecting and accepting the embedded Python predicate as arbitrary code
|
|
124
|
+
running with your user account's permissions.
|
|
125
|
+
|
|
126
|
+
`materialize` writes recorded values as deterministic JSON.
|
|
127
|
+
`reproduce-predicate` evaluates the embedded predicate in a fresh constrained
|
|
128
|
+
directory and reports every trial. Neither command is application replay.
|
|
129
|
+
|
|
130
|
+
The pre-0.1 `minimize` and `replay` names remain warning aliases for `reduce`
|
|
131
|
+
and `materialize`. They may be removed before 0.2.
|
|
132
|
+
|
|
133
|
+
Application replay is not part of the 0.1 CLI or promise. Experimental 0.5
|
|
134
|
+
library work now has one OpenAI Agents SDK adapter that re-executes an explicit
|
|
135
|
+
application callback through injected public `Model` and `FunctionTool`
|
|
136
|
+
interfaces. It enforces ordered exact matching and measured canaries, but
|
|
137
|
+
remains synthetic-only and is not a 0.5 readiness claim. See
|
|
138
|
+
[the application-replay boundary](docs/application-replay.md).
|
|
139
|
+
|
|
140
|
+
Predicate exit codes are strict:
|
|
141
|
+
|
|
142
|
+
- `0`: target failure reproduced.
|
|
143
|
+
- `1`: target failure absent.
|
|
144
|
+
- `2`: candidate or harness invalid.
|
|
145
|
+
- timeout, signal, cancellation, output overflow, unexpected exit, or missing
|
|
146
|
+
predicate file: invalid.
|
|
147
|
+
|
|
148
|
+
Invalid is never treated as absent and is never accepted as a reduction.
|
|
149
|
+
|
|
150
|
+
## Reproducible proof
|
|
151
|
+
|
|
152
|
+
The repository includes a synthetic 247-event mechanical fixture. This command builds it,
|
|
153
|
+
reduces it to at most 10 events, verifies 1-minimality independently, exports a
|
|
154
|
+
standalone reproduction, and runs it without an API key:
|
|
155
|
+
|
|
156
|
+
```bash
|
|
157
|
+
python scripts/killer_demo.py
|
|
158
|
+
```
|
|
159
|
+
|
|
160
|
+
The release gate requires this command to finish within 20 seconds. “1-minimal”
|
|
161
|
+
means deleting any remaining declared unit makes the failure absent or the
|
|
162
|
+
candidate structurally invalid. It does not mean globally smallest.
|
|
163
|
+
|
|
164
|
+
Two additional copy-paste checks:
|
|
165
|
+
|
|
166
|
+
```bash
|
|
167
|
+
python -m scripts.verify
|
|
168
|
+
python -m scripts.final_release_gate
|
|
169
|
+
```
|
|
170
|
+
|
|
171
|
+
The first runs tests, lint, typing, and contract self-tests. The second runs the
|
|
172
|
+
current exact-head security, secrets, killer-demo, and minimality checks from a
|
|
173
|
+
clean Git state. The immutable contract-v2 release gate remains historical
|
|
174
|
+
evidence, as documented in `RELEASING.md`.
|
|
175
|
+
|
|
176
|
+
## Scope and comparison
|
|
177
|
+
|
|
178
|
+
ReproSieve is not a trace viewer, observability backend, VM sandbox, or general
|
|
179
|
+
record/replay system. Existing record/replay tools preserve executions; ReproSieve
|
|
180
|
+
adds dependency-aware reduction, redaction before its own persistence, strict
|
|
181
|
+
tri-state predicates, and an independently checked 1-minimality claim for one
|
|
182
|
+
recorded agent trajectory. It does not diagnose root cause, rerun application
|
|
183
|
+
logic, or replay live model semantics.
|
|
184
|
+
|
|
185
|
+
## Privacy boundary
|
|
186
|
+
|
|
187
|
+
ReproSieve redacts SDK payloads in memory before its own first write. Capture
|
|
188
|
+
replaces the SDK default exporter unless `--retain-sdk-exporter` is explicitly
|
|
189
|
+
passed. Captured target stdout and stderr are discarded because they may contain
|
|
190
|
+
secrets. Exact canaries, bounded regexes, allow paths, deny paths, declared
|
|
191
|
+
workspace files, and environment allowlists are available on `capture`.
|
|
192
|
+
|
|
193
|
+
Arbitrary personal data cannot be detected reliably. Inspect every capsule
|
|
194
|
+
before publishing it. See [the privacy contract](docs/PRIVACY.md) and
|
|
195
|
+
[security review](docs/security-review.md).
|
|
196
|
+
|
|
197
|
+
## Development
|
|
198
|
+
|
|
199
|
+
```bash
|
|
200
|
+
python -m pip install -e ".[dev]"
|
|
201
|
+
python -m pytest
|
|
202
|
+
python -m ruff check .
|
|
203
|
+
python -m mypy
|
|
204
|
+
python -m build
|
|
205
|
+
python scripts/security_check.py
|
|
206
|
+
python scripts/killer_demo.py
|
|
207
|
+
```
|
|
208
|
+
|
|
209
|
+
Apache-2.0 licensed.
|
|
210
|
+
|
|
211
|
+
See [SUPPORT.md](SUPPORT.md) for supported combinations and
|
|
212
|
+
[CONTRIBUTING.md](CONTRIBUTING.md) before opening a change. Maturity levels are
|
|
213
|
+
defined in [ROADMAP.md](ROADMAP.md); real-case evidence status is tracked in
|
|
214
|
+
[docs/case-studies](docs/case-studies/README.md).
|
|
@@ -0,0 +1,179 @@
|
|
|
1
|
+
# ReproSieve
|
|
2
|
+
|
|
3
|
+
**Turn a large failed agent run into a small, redacted, offline reproduction.**
|
|
4
|
+
|
|
5
|
+
ReproSieve captures one supported OpenAI Agents SDK trace, removes secrets before
|
|
6
|
+
persistence, replaces model and tool calls with recorded outputs, and reduces the
|
|
7
|
+
trace against an executable failure predicate. The result is a deterministic,
|
|
8
|
+
hash-addressed capsule with an independent 1-minimality proof.
|
|
9
|
+
|
|
10
|
+
> **Status:** honest pre-0.1 seed. The end-to-end path works and has synthetic
|
|
11
|
+
> security fixtures, but it has not been proven safe for real credentials,
|
|
12
|
+
> private source, or personal data. Use synthetic or disposable inputs.
|
|
13
|
+
|
|
14
|
+
## Supported path
|
|
15
|
+
|
|
16
|
+
- Python 3.11, 3.12, and 3.13.
|
|
17
|
+
- `openai-agents>=0.18.3,<0.19`; CI tests 0.18.3.
|
|
18
|
+
- One completed SDK trace captured through the public `TracingProcessor` API.
|
|
19
|
+
- Embedded Python predicates. ReproSieve invokes them with a direct argument vector,
|
|
20
|
+
a clean temporary directory, provider keys removed, bounded time/output/process
|
|
21
|
+
resources, and Python audit hooks that deny outbound network, child processes,
|
|
22
|
+
native loading, and host-file access. These controls are defense in depth, not
|
|
23
|
+
an OS sandbox; embedded predicates are arbitrary code.
|
|
24
|
+
- Deterministic recorded-output materialization and offline predicate
|
|
25
|
+
reproduction. These paths reconstruct retained values and execute only the
|
|
26
|
+
declared predicate; they do not rerun application or orchestration code.
|
|
27
|
+
|
|
28
|
+
## Support matrix
|
|
29
|
+
|
|
30
|
+
| Surface | Supported |
|
|
31
|
+
|---|---|
|
|
32
|
+
| Python | 3.11, 3.12, 3.13 |
|
|
33
|
+
| Capture SDK | `openai-agents>=0.18.3,<0.19` |
|
|
34
|
+
| Standalone reproduction | Linux and macOS |
|
|
35
|
+
| Input safety claim | Synthetic or disposable data only |
|
|
36
|
+
|
|
37
|
+
Install the core:
|
|
38
|
+
|
|
39
|
+
```bash
|
|
40
|
+
python -m pip install reprosieve
|
|
41
|
+
```
|
|
42
|
+
|
|
43
|
+
Install capture support:
|
|
44
|
+
|
|
45
|
+
```bash
|
|
46
|
+
python -m pip install "reprosieve[openai]"
|
|
47
|
+
```
|
|
48
|
+
|
|
49
|
+
## Workflow
|
|
50
|
+
|
|
51
|
+
The predicate script is arbitrary capsule-provided Python and is not sandboxed.
|
|
52
|
+
It must be included in the capsule, and every command that can execute or export
|
|
53
|
+
it requires `--trust-embedded-predicate`. `--predicate` must be the final option
|
|
54
|
+
because everything after it is an argument vector.
|
|
55
|
+
|
|
56
|
+
```bash
|
|
57
|
+
reprosieve capture \
|
|
58
|
+
--output failed.reprosieve \
|
|
59
|
+
--workspace-root . \
|
|
60
|
+
--include verify_failure.py \
|
|
61
|
+
-- python app.py
|
|
62
|
+
|
|
63
|
+
reprosieve reduce failed.reprosieve \
|
|
64
|
+
--output-dir reduced \
|
|
65
|
+
--trust-embedded-predicate \
|
|
66
|
+
--predicate python verify_failure.py
|
|
67
|
+
|
|
68
|
+
reprosieve materialize reduced/<sha256>.reprosieve \
|
|
69
|
+
--output materialized.json
|
|
70
|
+
|
|
71
|
+
reprosieve reproduce-predicate reduced/<sha256>.reprosieve \
|
|
72
|
+
--trust-embedded-predicate \
|
|
73
|
+
--predicate python verify_failure.py
|
|
74
|
+
|
|
75
|
+
reprosieve verify-minimal reduced/<sha256>.reprosieve \
|
|
76
|
+
--trust-embedded-predicate \
|
|
77
|
+
--predicate python verify_failure.py
|
|
78
|
+
|
|
79
|
+
reprosieve export reduced/<sha256>.reprosieve \
|
|
80
|
+
--trust-embedded-predicate \
|
|
81
|
+
--output issue-repro
|
|
82
|
+
|
|
83
|
+
cd issue-repro
|
|
84
|
+
python reproduce.py --trust-embedded-predicate
|
|
85
|
+
```
|
|
86
|
+
|
|
87
|
+
Exported reproductions refuse to execute without that explicit flag. Use it only
|
|
88
|
+
after inspecting and accepting the embedded Python predicate as arbitrary code
|
|
89
|
+
running with your user account's permissions.
|
|
90
|
+
|
|
91
|
+
`materialize` writes recorded values as deterministic JSON.
|
|
92
|
+
`reproduce-predicate` evaluates the embedded predicate in a fresh constrained
|
|
93
|
+
directory and reports every trial. Neither command is application replay.
|
|
94
|
+
|
|
95
|
+
The pre-0.1 `minimize` and `replay` names remain warning aliases for `reduce`
|
|
96
|
+
and `materialize`. They may be removed before 0.2.
|
|
97
|
+
|
|
98
|
+
Application replay is not part of the 0.1 CLI or promise. Experimental 0.5
|
|
99
|
+
library work now has one OpenAI Agents SDK adapter that re-executes an explicit
|
|
100
|
+
application callback through injected public `Model` and `FunctionTool`
|
|
101
|
+
interfaces. It enforces ordered exact matching and measured canaries, but
|
|
102
|
+
remains synthetic-only and is not a 0.5 readiness claim. See
|
|
103
|
+
[the application-replay boundary](docs/application-replay.md).
|
|
104
|
+
|
|
105
|
+
Predicate exit codes are strict:
|
|
106
|
+
|
|
107
|
+
- `0`: target failure reproduced.
|
|
108
|
+
- `1`: target failure absent.
|
|
109
|
+
- `2`: candidate or harness invalid.
|
|
110
|
+
- timeout, signal, cancellation, output overflow, unexpected exit, or missing
|
|
111
|
+
predicate file: invalid.
|
|
112
|
+
|
|
113
|
+
Invalid is never treated as absent and is never accepted as a reduction.
|
|
114
|
+
|
|
115
|
+
## Reproducible proof
|
|
116
|
+
|
|
117
|
+
The repository includes a synthetic 247-event mechanical fixture. This command builds it,
|
|
118
|
+
reduces it to at most 10 events, verifies 1-minimality independently, exports a
|
|
119
|
+
standalone reproduction, and runs it without an API key:
|
|
120
|
+
|
|
121
|
+
```bash
|
|
122
|
+
python scripts/killer_demo.py
|
|
123
|
+
```
|
|
124
|
+
|
|
125
|
+
The release gate requires this command to finish within 20 seconds. “1-minimal”
|
|
126
|
+
means deleting any remaining declared unit makes the failure absent or the
|
|
127
|
+
candidate structurally invalid. It does not mean globally smallest.
|
|
128
|
+
|
|
129
|
+
Two additional copy-paste checks:
|
|
130
|
+
|
|
131
|
+
```bash
|
|
132
|
+
python -m scripts.verify
|
|
133
|
+
python -m scripts.final_release_gate
|
|
134
|
+
```
|
|
135
|
+
|
|
136
|
+
The first runs tests, lint, typing, and contract self-tests. The second runs the
|
|
137
|
+
current exact-head security, secrets, killer-demo, and minimality checks from a
|
|
138
|
+
clean Git state. The immutable contract-v2 release gate remains historical
|
|
139
|
+
evidence, as documented in `RELEASING.md`.
|
|
140
|
+
|
|
141
|
+
## Scope and comparison
|
|
142
|
+
|
|
143
|
+
ReproSieve is not a trace viewer, observability backend, VM sandbox, or general
|
|
144
|
+
record/replay system. Existing record/replay tools preserve executions; ReproSieve
|
|
145
|
+
adds dependency-aware reduction, redaction before its own persistence, strict
|
|
146
|
+
tri-state predicates, and an independently checked 1-minimality claim for one
|
|
147
|
+
recorded agent trajectory. It does not diagnose root cause, rerun application
|
|
148
|
+
logic, or replay live model semantics.
|
|
149
|
+
|
|
150
|
+
## Privacy boundary
|
|
151
|
+
|
|
152
|
+
ReproSieve redacts SDK payloads in memory before its own first write. Capture
|
|
153
|
+
replaces the SDK default exporter unless `--retain-sdk-exporter` is explicitly
|
|
154
|
+
passed. Captured target stdout and stderr are discarded because they may contain
|
|
155
|
+
secrets. Exact canaries, bounded regexes, allow paths, deny paths, declared
|
|
156
|
+
workspace files, and environment allowlists are available on `capture`.
|
|
157
|
+
|
|
158
|
+
Arbitrary personal data cannot be detected reliably. Inspect every capsule
|
|
159
|
+
before publishing it. See [the privacy contract](docs/PRIVACY.md) and
|
|
160
|
+
[security review](docs/security-review.md).
|
|
161
|
+
|
|
162
|
+
## Development
|
|
163
|
+
|
|
164
|
+
```bash
|
|
165
|
+
python -m pip install -e ".[dev]"
|
|
166
|
+
python -m pytest
|
|
167
|
+
python -m ruff check .
|
|
168
|
+
python -m mypy
|
|
169
|
+
python -m build
|
|
170
|
+
python scripts/security_check.py
|
|
171
|
+
python scripts/killer_demo.py
|
|
172
|
+
```
|
|
173
|
+
|
|
174
|
+
Apache-2.0 licensed.
|
|
175
|
+
|
|
176
|
+
See [SUPPORT.md](SUPPORT.md) for supported combinations and
|
|
177
|
+
[CONTRIBUTING.md](CONTRIBUTING.md) before opening a change. Maturity levels are
|
|
178
|
+
defined in [ROADMAP.md](ROADMAP.md); real-case evidence status is tracked in
|
|
179
|
+
[docs/case-studies](docs/case-studies/README.md).
|