@mnemom/agent-alignment-protocol 0.1.1 → 0.1.3
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.
- package/LICENSE +191 -0
- package/README.md +333 -0
- package/dist/index.d.mts +1 -1
- package/dist/index.d.ts +1 -1
- package/dist/index.js +2 -10
- package/dist/index.mjs +2 -10
- package/package.json +6 -4
- package/src/constants.ts +1 -1
- package/src/verification/api.ts +2 -2
- package/src/verification/features.ts +4 -10
package/LICENSE
ADDED
|
@@ -0,0 +1,191 @@
|
|
|
1
|
+
Apache License
|
|
2
|
+
Version 2.0, January 2004
|
|
3
|
+
http://www.apache.org/licenses/
|
|
4
|
+
|
|
5
|
+
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
|
|
6
|
+
|
|
7
|
+
1. Definitions.
|
|
8
|
+
|
|
9
|
+
"License" shall mean the terms and conditions for use, reproduction,
|
|
10
|
+
and distribution as defined by Sections 1 through 9 of this document.
|
|
11
|
+
|
|
12
|
+
"Licensor" shall mean the copyright owner or entity authorized by
|
|
13
|
+
the copyright owner that is granting the License.
|
|
14
|
+
|
|
15
|
+
"Legal Entity" shall mean the union of the acting entity and all
|
|
16
|
+
other entities that control, are controlled by, or are under common
|
|
17
|
+
control with that entity. For the purposes of this definition,
|
|
18
|
+
"control" means (i) the power, direct or indirect, to cause the
|
|
19
|
+
direction or management of such entity, whether by contract or
|
|
20
|
+
otherwise, or (ii) ownership of fifty percent (50%) or more of the
|
|
21
|
+
outstanding shares, or (iii) beneficial ownership of such entity.
|
|
22
|
+
|
|
23
|
+
"You" (or "Your") shall mean an individual or Legal Entity
|
|
24
|
+
exercising permissions granted by this License.
|
|
25
|
+
|
|
26
|
+
"Source" form shall mean the preferred form for making modifications,
|
|
27
|
+
including but not limited to software source code, documentation
|
|
28
|
+
source, and configuration files.
|
|
29
|
+
|
|
30
|
+
"Object" form shall mean any form resulting from mechanical
|
|
31
|
+
transformation or translation of a Source form, including but
|
|
32
|
+
not limited to compiled object code, generated documentation,
|
|
33
|
+
and conversions to other media types.
|
|
34
|
+
|
|
35
|
+
"Work" shall mean the work of authorship, whether in Source or
|
|
36
|
+
Object form, made available under the License, as indicated by a
|
|
37
|
+
copyright notice that is included in or attached to the work
|
|
38
|
+
(an example is provided in the Appendix below).
|
|
39
|
+
|
|
40
|
+
"Derivative Works" shall mean any work, whether in Source or Object
|
|
41
|
+
form, that is based on (or derived from) the Work and for which the
|
|
42
|
+
editorial revisions, annotations, elaborations, or other modifications
|
|
43
|
+
represent, as a whole, an original work of authorship. For the purposes
|
|
44
|
+
of this License, Derivative Works shall not include works that remain
|
|
45
|
+
separable from, or merely link (or bind by name) to the interfaces of,
|
|
46
|
+
the Work and Derivative Works thereof.
|
|
47
|
+
|
|
48
|
+
"Contribution" shall mean any work of authorship, including
|
|
49
|
+
the original version of the Work and any modifications or additions
|
|
50
|
+
to that Work or Derivative Works thereof, that is intentionally
|
|
51
|
+
submitted to the Licensor for inclusion in the Work by the copyright
|
|
52
|
+
owner or by an individual or Legal Entity authorized to submit on
|
|
53
|
+
behalf of the copyright owner. For the purposes of this definition,
|
|
54
|
+
"submitted" means any form of electronic, verbal, or written
|
|
55
|
+
communication sent to the Licensor or its representatives, including
|
|
56
|
+
but not limited to communication on electronic mailing lists, source
|
|
57
|
+
code control systems, and issue tracking systems that are managed by,
|
|
58
|
+
or on behalf of, the Licensor for the purpose of discussing and
|
|
59
|
+
improving the Work, but excluding communication that is conspicuously
|
|
60
|
+
marked or otherwise designated in writing by the copyright owner as
|
|
61
|
+
"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
|
+
Copyright 2026 Mnemom LLC
|
|
180
|
+
|
|
181
|
+
Licensed under the Apache License, Version 2.0 (the "License");
|
|
182
|
+
you may not use this file except in compliance with the License.
|
|
183
|
+
You may obtain a copy of the License at
|
|
184
|
+
|
|
185
|
+
http://www.apache.org/licenses/LICENSE-2.0
|
|
186
|
+
|
|
187
|
+
Unless required by applicable law or agreed to in writing, software
|
|
188
|
+
distributed under the License is distributed on an "AS IS" BASIS,
|
|
189
|
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
190
|
+
See the License for the specific language governing permissions and
|
|
191
|
+
limitations under the License.
|
package/README.md
ADDED
|
@@ -0,0 +1,333 @@
|
|
|
1
|
+
# Agent Alignment Protocol (AAP)
|
|
2
|
+
|
|
3
|
+
[](https://github.com/mnemom/aap/actions/workflows/ci.yml)
|
|
4
|
+
[](https://github.com/mnemom/aap/actions/workflows/codeql.yml)
|
|
5
|
+
[](https://codecov.io/gh/mnemom/aap)
|
|
6
|
+
[](https://pypi.org/project/agent-alignment-protocol/)
|
|
7
|
+
[](https://www.npmjs.com/package/@mnemom/agent-alignment-protocol)
|
|
8
|
+
[](LICENSE)
|
|
9
|
+
[](docs/SPEC.md)
|
|
10
|
+
|
|
11
|
+
**A transparency protocol for autonomous agents.**
|
|
12
|
+
|
|
13
|
+
AAP lets agents declare their alignment posture, produce auditable decision traces, and verify value coherence before coordinating with other agents. It extends existing protocols (A2A, MCP) with an alignment layer that makes agent behavior observable.
|
|
14
|
+
|
|
15
|
+
> AAP is a transparency protocol, not a trust protocol. It makes agent behavior more observable, not more guaranteed.
|
|
16
|
+
|
|
17
|
+
## Quick Start
|
|
18
|
+
|
|
19
|
+
```bash
|
|
20
|
+
# Install
|
|
21
|
+
pip install agent-alignment-protocol
|
|
22
|
+
|
|
23
|
+
# Generate an Alignment Card
|
|
24
|
+
aap init --values "principal_benefit,transparency,harm_prevention"
|
|
25
|
+
# ✓ Created alignment-card.json
|
|
26
|
+
|
|
27
|
+
# Instrument your agent
|
|
28
|
+
```
|
|
29
|
+
|
|
30
|
+
```python
|
|
31
|
+
from aap import trace_decision
|
|
32
|
+
|
|
33
|
+
@trace_decision(card_path="alignment-card.json")
|
|
34
|
+
def recommend_product(user_preferences):
|
|
35
|
+
# Your agent logic here
|
|
36
|
+
# Decisions are automatically traced
|
|
37
|
+
...
|
|
38
|
+
```
|
|
39
|
+
|
|
40
|
+
```bash
|
|
41
|
+
# Verify behavior matches declaration
|
|
42
|
+
aap verify --card alignment-card.json --trace logs/trace.json
|
|
43
|
+
# ✓ Verified [similarity: 0.82]
|
|
44
|
+
# Checks: autonomy, escalation, values, forbidden, behavioral_similarity
|
|
45
|
+
```
|
|
46
|
+
|
|
47
|
+
## Why AAP?
|
|
48
|
+
|
|
49
|
+
The agent protocol stack provides capability discovery (A2A), tool integration (MCP), and payment authorization (AP2). None address a fundamental question: **Is this agent serving its principal's interests?**
|
|
50
|
+
|
|
51
|
+
| Protocol | Function | Gap |
|
|
52
|
+
|----------|----------|-----|
|
|
53
|
+
| **MCP** | Agent-to-tool connectivity | No alignment semantics |
|
|
54
|
+
| **A2A** | Task negotiation between agents | No value verification |
|
|
55
|
+
| **AP2** | Payment authorization | No behavioral audit |
|
|
56
|
+
|
|
57
|
+
As agent capabilities become symmetric—equal access to information, equal reasoning power—alignment becomes the primary differentiator. AAP provides the infrastructure to make alignment claims verifiable.
|
|
58
|
+
|
|
59
|
+
## Three Components
|
|
60
|
+
|
|
61
|
+
```
|
|
62
|
+
┌─────────────────┬─────────────────┬─────────────────┐
|
|
63
|
+
│ Alignment Card │ AP-Trace │ Value Coherence │
|
|
64
|
+
│ │ │ Handshake │
|
|
65
|
+
├─────────────────┼─────────────────┼─────────────────┤
|
|
66
|
+
│ "What I claim │ "What I │ "Can we work │
|
|
67
|
+
│ to be" │ actually did" │ together?" │
|
|
68
|
+
└─────────────────┴─────────────────┴─────────────────┘
|
|
69
|
+
Declaration Audit Coordination
|
|
70
|
+
```
|
|
71
|
+
|
|
72
|
+
### Alignment Card
|
|
73
|
+
|
|
74
|
+
A structured declaration of an agent's alignment posture:
|
|
75
|
+
|
|
76
|
+
```json
|
|
77
|
+
{
|
|
78
|
+
"aap_version": "0.1.0",
|
|
79
|
+
"agent_id": "did:web:my-agent.example.com",
|
|
80
|
+
"principal": {
|
|
81
|
+
"type": "human",
|
|
82
|
+
"relationship": "delegated_authority"
|
|
83
|
+
},
|
|
84
|
+
"values": {
|
|
85
|
+
"declared": ["principal_benefit", "transparency", "minimal_data"],
|
|
86
|
+
"conflicts_with": ["deceptive_marketing", "hidden_fees"]
|
|
87
|
+
},
|
|
88
|
+
"autonomy_envelope": {
|
|
89
|
+
"bounded_actions": ["search", "compare", "recommend"],
|
|
90
|
+
"escalation_triggers": [
|
|
91
|
+
{
|
|
92
|
+
"condition": "purchase_value > 100",
|
|
93
|
+
"action": "escalate",
|
|
94
|
+
"reason": "Exceeds autonomous spending limit"
|
|
95
|
+
}
|
|
96
|
+
],
|
|
97
|
+
"forbidden_actions": ["share_credentials", "subscribe_to_services"]
|
|
98
|
+
},
|
|
99
|
+
"audit_commitment": {
|
|
100
|
+
"trace_format": "ap-trace-v1",
|
|
101
|
+
"retention_days": 90,
|
|
102
|
+
"queryable": true
|
|
103
|
+
}
|
|
104
|
+
}
|
|
105
|
+
```
|
|
106
|
+
|
|
107
|
+
### AP-Trace
|
|
108
|
+
|
|
109
|
+
An audit log entry recording each decision:
|
|
110
|
+
|
|
111
|
+
```json
|
|
112
|
+
{
|
|
113
|
+
"trace_id": "tr-f47ac10b-58cc-4372",
|
|
114
|
+
"card_id": "ac-f47ac10b-58cc-4372",
|
|
115
|
+
"timestamp": "2026-01-31T12:30:00Z",
|
|
116
|
+
"action": {
|
|
117
|
+
"type": "recommend",
|
|
118
|
+
"name": "product_recommendation",
|
|
119
|
+
"category": "bounded"
|
|
120
|
+
},
|
|
121
|
+
"decision": {
|
|
122
|
+
"alternatives_considered": [
|
|
123
|
+
{"option_id": "A", "score": 0.85, "flags": []},
|
|
124
|
+
{"option_id": "B", "score": 0.72, "flags": ["sponsored_content"]}
|
|
125
|
+
],
|
|
126
|
+
"selected": "A",
|
|
127
|
+
"selection_reasoning": "Highest score. Option B flagged as sponsored and deprioritized per principal_benefit value.",
|
|
128
|
+
"values_applied": ["principal_benefit", "transparency"]
|
|
129
|
+
},
|
|
130
|
+
"escalation": {
|
|
131
|
+
"evaluated": true,
|
|
132
|
+
"required": false,
|
|
133
|
+
"reason": "Recommendation only, no purchase action"
|
|
134
|
+
}
|
|
135
|
+
}
|
|
136
|
+
```
|
|
137
|
+
|
|
138
|
+
### Value Coherence Handshake
|
|
139
|
+
|
|
140
|
+
Pre-coordination compatibility check between agents:
|
|
141
|
+
|
|
142
|
+
```python
|
|
143
|
+
from aap import check_coherence
|
|
144
|
+
|
|
145
|
+
result = check_coherence(my_card, their_card, task_context)
|
|
146
|
+
|
|
147
|
+
if result.compatible:
|
|
148
|
+
# Proceed with coordination
|
|
149
|
+
proceed_with_task()
|
|
150
|
+
else:
|
|
151
|
+
# Handle conflict
|
|
152
|
+
print(f"Value conflict: {result.conflicts}")
|
|
153
|
+
# Escalate to principals or negotiate scope
|
|
154
|
+
```
|
|
155
|
+
|
|
156
|
+
## What AAP Does Not Do
|
|
157
|
+
|
|
158
|
+
This matters. Read it.
|
|
159
|
+
|
|
160
|
+
1. **AAP does NOT ensure alignment—it provides visibility.** An agent can produce perfect traces while acting against its principal's interests.
|
|
161
|
+
|
|
162
|
+
2. **Verified does NOT equal safe.** A verified trace means consistency with declared alignment. It doesn't mean the alignment is good or the outcome was beneficial.
|
|
163
|
+
|
|
164
|
+
3. **AP-Trace is sampled, not complete.** Traces capture decision points, not every computation. Significant reasoning may occur between traces.
|
|
165
|
+
|
|
166
|
+
4. **Value coherence is relative to declared values.** The handshake checks if declared values are compatible. It doesn't verify agents hold these values or will act on them.
|
|
167
|
+
|
|
168
|
+
5. **Tested on transformer-based agents.** Other architectures may exhibit behaviors AAP doesn't capture.
|
|
169
|
+
|
|
170
|
+
For the complete limitations disclosure, see [Section 10 of the Specification](docs/SPEC.md#10-limitations).
|
|
171
|
+
|
|
172
|
+
## Installation
|
|
173
|
+
|
|
174
|
+
```bash
|
|
175
|
+
# Python
|
|
176
|
+
pip install agent-alignment-protocol
|
|
177
|
+
|
|
178
|
+
# TypeScript
|
|
179
|
+
npm install agent-alignment-protocol
|
|
180
|
+
```
|
|
181
|
+
|
|
182
|
+
## Integration
|
|
183
|
+
|
|
184
|
+
### With A2A Agents
|
|
185
|
+
|
|
186
|
+
AAP extends the A2A Agent Card with an `alignment` block:
|
|
187
|
+
|
|
188
|
+
```json
|
|
189
|
+
{
|
|
190
|
+
"name": "Shopping Assistant",
|
|
191
|
+
"description": "Helps users find products",
|
|
192
|
+
"url": "https://shopping.example.com",
|
|
193
|
+
"alignment": {
|
|
194
|
+
"$ref": "./alignment-card.json"
|
|
195
|
+
}
|
|
196
|
+
}
|
|
197
|
+
```
|
|
198
|
+
|
|
199
|
+
See [A2A Migration Guide](docs/a2a-migration.md).
|
|
200
|
+
|
|
201
|
+
### With MCP Tools
|
|
202
|
+
|
|
203
|
+
AAP can trace tool invocations and verify they match declared alignment:
|
|
204
|
+
|
|
205
|
+
```python
|
|
206
|
+
from aap import mcp_traced
|
|
207
|
+
|
|
208
|
+
@mcp_traced(card_path="alignment-card.json")
|
|
209
|
+
def my_tool(params):
|
|
210
|
+
...
|
|
211
|
+
```
|
|
212
|
+
|
|
213
|
+
See [MCP Migration Guide](docs/mcp-migration.md).
|
|
214
|
+
|
|
215
|
+
## Verification
|
|
216
|
+
|
|
217
|
+
The verification engine checks whether observed behavior matches declared alignment:
|
|
218
|
+
|
|
219
|
+
```python
|
|
220
|
+
from aap import verify_trace, detect_drift
|
|
221
|
+
|
|
222
|
+
# Single trace verification
|
|
223
|
+
result = verify_trace(trace, card)
|
|
224
|
+
print(f"Verified: {result.verified}, Similarity: {result.similarity_score:.2f}")
|
|
225
|
+
if not result.verified:
|
|
226
|
+
print(f"Violations: {result.violations}")
|
|
227
|
+
if result.warnings:
|
|
228
|
+
print(f"Warnings: {result.warnings}")
|
|
229
|
+
|
|
230
|
+
# Drift detection over time
|
|
231
|
+
alerts = detect_drift(card, recent_traces)
|
|
232
|
+
for alert in alerts:
|
|
233
|
+
print(f"Drift detected: {alert.analysis.drift_direction}")
|
|
234
|
+
```
|
|
235
|
+
|
|
236
|
+
**Verification checks:**
|
|
237
|
+
- Autonomy compliance (actions within declared bounds)
|
|
238
|
+
- Escalation compliance (required escalations were performed)
|
|
239
|
+
- Value consistency (applied values match declared values)
|
|
240
|
+
- Forbidden action compliance (no forbidden actions taken)
|
|
241
|
+
- Behavioral similarity (semantic alignment using SSM analysis)
|
|
242
|
+
|
|
243
|
+
**Similarity scoring:** Each verification returns a `similarity_score` (0.0-1.0) measuring semantic similarity between the trace and declared alignment. If a trace passes structural checks but has `similarity_score < 0.50`, a `low_behavioral_similarity` warning is generated.
|
|
244
|
+
|
|
245
|
+
## Try It
|
|
246
|
+
|
|
247
|
+
**[Interactive Playground](https://mnemom.github.io/aap/playground/)** — Verify traces in your browser with SSM visualization.
|
|
248
|
+
|
|
249
|
+
- Paste your Alignment Card and AP-Trace
|
|
250
|
+
- See verification results with similarity scoring
|
|
251
|
+
- Visualize behavioral patterns with SSM heatmaps
|
|
252
|
+
- Adjust thresholds in real-time
|
|
253
|
+
|
|
254
|
+
No server required — runs entirely client-side via WebAssembly.
|
|
255
|
+
|
|
256
|
+
## Documentation
|
|
257
|
+
|
|
258
|
+
| Document | Description |
|
|
259
|
+
|----------|-------------|
|
|
260
|
+
| [**SPEC.md**](docs/SPEC.md) | Full protocol specification (IETF-style) |
|
|
261
|
+
| [**QUICKSTART.md**](docs/QUICKSTART.md) | Zero to compliant in 5 minutes |
|
|
262
|
+
| [**LIMITS.md**](docs/LIMITS.md) | What AAP guarantees and doesn't |
|
|
263
|
+
| [**SECURITY.md**](docs/SECURITY.md) | Threat model and security considerations |
|
|
264
|
+
| [**CALIBRATION.md**](docs/CALIBRATION.md) | How verification thresholds were derived |
|
|
265
|
+
|
|
266
|
+
## Examples
|
|
267
|
+
|
|
268
|
+
| Example | Description |
|
|
269
|
+
|---------|-------------|
|
|
270
|
+
| [`simple-agent/`](examples/simple-agent/) | Minimal AAP implementation |
|
|
271
|
+
| [`a2a-integration/`](examples/a2a-integration/) | A2A agent with AAP |
|
|
272
|
+
| [`mcp-integration/`](examples/mcp-integration/) | MCP tools with alignment |
|
|
273
|
+
| [`alignment-failure/`](examples/alignment-failure/) | Deliberate failure for testing |
|
|
274
|
+
|
|
275
|
+
## Status
|
|
276
|
+
|
|
277
|
+
**Current Version**: 0.1.1 (Draft)
|
|
278
|
+
|
|
279
|
+
| Component | Status |
|
|
280
|
+
|-----------|--------|
|
|
281
|
+
| Specification | ✅ Complete |
|
|
282
|
+
| JSON Schemas | ✅ Complete |
|
|
283
|
+
| Python SDK | ✅ Complete |
|
|
284
|
+
| TypeScript SDK | ✅ Complete |
|
|
285
|
+
| Verification Engine | ✅ Complete (with similarity scoring) |
|
|
286
|
+
| SSM Visualization | ✅ Complete |
|
|
287
|
+
| Interactive Playground | ✅ Complete |
|
|
288
|
+
|
|
289
|
+
## API Reference
|
|
290
|
+
|
|
291
|
+
```python
|
|
292
|
+
# Core API
|
|
293
|
+
from aap import (
|
|
294
|
+
verify_trace, # Verify single trace against card → VerificationResult
|
|
295
|
+
check_coherence, # Check value compatibility between agents → CoherenceResult
|
|
296
|
+
detect_drift, # Detect behavioral drift over time → list[DriftAlert]
|
|
297
|
+
trace_decision, # Decorator for automatic AP-Trace generation
|
|
298
|
+
mcp_traced, # Decorator for MCP tool tracing
|
|
299
|
+
)
|
|
300
|
+
|
|
301
|
+
# Models
|
|
302
|
+
from aap import (
|
|
303
|
+
AlignmentCard,
|
|
304
|
+
APTrace,
|
|
305
|
+
VerificationResult, # .verified, .similarity_score, .violations, .warnings
|
|
306
|
+
CoherenceResult, # .compatible, .score, .value_alignment
|
|
307
|
+
DriftAlert, # .analysis.similarity_score, .analysis.drift_direction
|
|
308
|
+
)
|
|
309
|
+
|
|
310
|
+
# CLI
|
|
311
|
+
# aap init [--values VALUES] [--output FILE]
|
|
312
|
+
# aap verify --card CARD --trace TRACE → Shows [similarity: X.XX]
|
|
313
|
+
# aap check-coherence --my-card MINE --their-card THEIRS
|
|
314
|
+
# aap drift --card CARD --traces TRACES_DIR → Uses SSM analysis
|
|
315
|
+
```
|
|
316
|
+
|
|
317
|
+
## Contributing
|
|
318
|
+
|
|
319
|
+
We welcome contributions. See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
|
|
320
|
+
|
|
321
|
+
Key areas where we need help:
|
|
322
|
+
- SDK implementations in other languages
|
|
323
|
+
- Integration examples with popular agent frameworks
|
|
324
|
+
- Test vectors for edge cases
|
|
325
|
+
- Documentation improvements
|
|
326
|
+
|
|
327
|
+
## License
|
|
328
|
+
|
|
329
|
+
Apache 2.0. See [LICENSE](LICENSE) for details.
|
|
330
|
+
|
|
331
|
+
---
|
|
332
|
+
|
|
333
|
+
*Agent Alignment Protocol — Making agent alignment observable.*
|
package/dist/index.d.mts
CHANGED
|
@@ -678,6 +678,6 @@ declare const CONFLICT_PENALTY_MULTIPLIER = 0.5;
|
|
|
678
678
|
declare const MIN_WORD_LENGTH = 3;
|
|
679
679
|
/** Maximum features to extract from TF-IDF vectorization */
|
|
680
680
|
declare const MAX_TFIDF_FEATURES = 500;
|
|
681
|
-
declare const ALGORITHM_VERSION = "1.
|
|
681
|
+
declare const ALGORITHM_VERSION = "1.1.0";
|
|
682
682
|
|
|
683
683
|
export { ALGORITHM_VERSION, type APTrace, type Action, type ActionCategory, type ActionTarget, type ActionType, type AlignmentCard, type AlignmentCardRequest, type AlignmentCardResponse, type Alternative, type AuditCommitment, type AuditStorage, type AutonomyEnvelope, type AutonomyScope, CONFLICT_PENALTY_MULTIPLIER, type Coherence, type CoherenceResult, type CoherenceResultMessage, DEFAULT_SIMILARITY_THRESHOLD, DEFAULT_SUSTAINED_TURNS_THRESHOLD, type DataSharing, type Decision, type DriftAlert, type DriftAnalysis, type DriftDirection, type DriftIndicator, type Escalation, type EscalationStatus, type EscalationTrigger, type HierarchyType, MAX_TFIDF_FEATURES, MIN_COHERENCE_FOR_PROCEED, MIN_WORD_LENGTH, type MonetaryValue, NEAR_BOUNDARY_THRESHOLD, type Principal, type PrincipalResponse, type PrincipalType, type ProposedCollaboration, type ProposedResolution, type RelationshipType, type RequesterInfo, type Severity, type Signature, type StorageType, type TamperEvidence, type TaskContext, type TraceContext, type TriggerAction, type TriggerCheck, VIOLATION_SEVERITY, type ValueAlignment, type ValueAlignmentDetail, type ValueCoherenceCheck, type ValueCoherenceMessage, type ValueConflict, type ValueConflictResult, type ValueDefinition, type Values, type VerificationMetadata, type VerificationResult, type Violation, type ViolationType, type Warning, checkCoherence, cosineSimilarity, createViolation, detectDrift, extractCardFeatures, extractTraceFeatures, getSelectedAlternative, hadViolations, hasValue, isActionBounded, isActionForbidden, isCardExpired, verifyTrace, wasEscalated };
|
package/dist/index.d.ts
CHANGED
|
@@ -678,6 +678,6 @@ declare const CONFLICT_PENALTY_MULTIPLIER = 0.5;
|
|
|
678
678
|
declare const MIN_WORD_LENGTH = 3;
|
|
679
679
|
/** Maximum features to extract from TF-IDF vectorization */
|
|
680
680
|
declare const MAX_TFIDF_FEATURES = 500;
|
|
681
|
-
declare const ALGORITHM_VERSION = "1.
|
|
681
|
+
declare const ALGORITHM_VERSION = "1.1.0";
|
|
682
682
|
|
|
683
683
|
export { ALGORITHM_VERSION, type APTrace, type Action, type ActionCategory, type ActionTarget, type ActionType, type AlignmentCard, type AlignmentCardRequest, type AlignmentCardResponse, type Alternative, type AuditCommitment, type AuditStorage, type AutonomyEnvelope, type AutonomyScope, CONFLICT_PENALTY_MULTIPLIER, type Coherence, type CoherenceResult, type CoherenceResultMessage, DEFAULT_SIMILARITY_THRESHOLD, DEFAULT_SUSTAINED_TURNS_THRESHOLD, type DataSharing, type Decision, type DriftAlert, type DriftAnalysis, type DriftDirection, type DriftIndicator, type Escalation, type EscalationStatus, type EscalationTrigger, type HierarchyType, MAX_TFIDF_FEATURES, MIN_COHERENCE_FOR_PROCEED, MIN_WORD_LENGTH, type MonetaryValue, NEAR_BOUNDARY_THRESHOLD, type Principal, type PrincipalResponse, type PrincipalType, type ProposedCollaboration, type ProposedResolution, type RelationshipType, type RequesterInfo, type Severity, type Signature, type StorageType, type TamperEvidence, type TaskContext, type TraceContext, type TriggerAction, type TriggerCheck, VIOLATION_SEVERITY, type ValueAlignment, type ValueAlignmentDetail, type ValueCoherenceCheck, type ValueCoherenceMessage, type ValueConflict, type ValueConflictResult, type ValueDefinition, type Values, type VerificationMetadata, type VerificationResult, type Violation, type ViolationType, type Warning, checkCoherence, cosineSimilarity, createViolation, detectDrift, extractCardFeatures, extractTraceFeatures, getSelectedAlternative, hadViolations, hasValue, isActionBounded, isActionForbidden, isCardExpired, verifyTrace, wasEscalated };
|
package/dist/index.js
CHANGED
|
@@ -54,7 +54,7 @@ var MIN_COHERENCE_FOR_PROCEED = 0.7;
|
|
|
54
54
|
var CONFLICT_PENALTY_MULTIPLIER = 0.5;
|
|
55
55
|
var MIN_WORD_LENGTH = 3;
|
|
56
56
|
var MAX_TFIDF_FEATURES = 500;
|
|
57
|
-
var ALGORITHM_VERSION = "1.
|
|
57
|
+
var ALGORITHM_VERSION = "1.1.0";
|
|
58
58
|
|
|
59
59
|
// src/verification/features.ts
|
|
60
60
|
var STOPWORDS = /* @__PURE__ */ new Set([
|
|
@@ -148,15 +148,7 @@ function extractTraceFeatures(trace) {
|
|
|
148
148
|
features[`escalation:${trace.escalation.escalation_status}`] = 1;
|
|
149
149
|
}
|
|
150
150
|
}
|
|
151
|
-
const reasoningTokens = tokenize(trace.decision.selection_reasoning);
|
|
152
|
-
for (const token of reasoningTokens) {
|
|
153
|
-
features[`content:${token}`] = (features[`content:${token}`] ?? 0) + 0.5;
|
|
154
|
-
}
|
|
155
151
|
for (const alt of trace.decision.alternatives_considered) {
|
|
156
|
-
const altTokens = tokenize(alt.description);
|
|
157
|
-
for (const token of altTokens) {
|
|
158
|
-
features[`content:${token}`] = (features[`content:${token}`] ?? 0) + 0.25;
|
|
159
|
-
}
|
|
160
152
|
if (alt.flags) {
|
|
161
153
|
for (const flag of alt.flags) {
|
|
162
154
|
features[`flag:${flag}`] = 1;
|
|
@@ -429,7 +421,7 @@ function detectDrift(card, traces, similarityThreshold = DEFAULT_SIMILARITY_THRE
|
|
|
429
421
|
} else {
|
|
430
422
|
lowSimilarityStreak = [];
|
|
431
423
|
}
|
|
432
|
-
if (lowSimilarityStreak.length
|
|
424
|
+
if (lowSimilarityStreak.length === sustainedThreshold) {
|
|
433
425
|
const latest = lowSimilarityStreak[lowSimilarityStreak.length - 1];
|
|
434
426
|
const direction = inferDriftDirection(
|
|
435
427
|
lowSimilarityStreak,
|
package/dist/index.mjs
CHANGED
|
@@ -6,7 +6,7 @@ var MIN_COHERENCE_FOR_PROCEED = 0.7;
|
|
|
6
6
|
var CONFLICT_PENALTY_MULTIPLIER = 0.5;
|
|
7
7
|
var MIN_WORD_LENGTH = 3;
|
|
8
8
|
var MAX_TFIDF_FEATURES = 500;
|
|
9
|
-
var ALGORITHM_VERSION = "1.
|
|
9
|
+
var ALGORITHM_VERSION = "1.1.0";
|
|
10
10
|
|
|
11
11
|
// src/verification/features.ts
|
|
12
12
|
var STOPWORDS = /* @__PURE__ */ new Set([
|
|
@@ -100,15 +100,7 @@ function extractTraceFeatures(trace) {
|
|
|
100
100
|
features[`escalation:${trace.escalation.escalation_status}`] = 1;
|
|
101
101
|
}
|
|
102
102
|
}
|
|
103
|
-
const reasoningTokens = tokenize(trace.decision.selection_reasoning);
|
|
104
|
-
for (const token of reasoningTokens) {
|
|
105
|
-
features[`content:${token}`] = (features[`content:${token}`] ?? 0) + 0.5;
|
|
106
|
-
}
|
|
107
103
|
for (const alt of trace.decision.alternatives_considered) {
|
|
108
|
-
const altTokens = tokenize(alt.description);
|
|
109
|
-
for (const token of altTokens) {
|
|
110
|
-
features[`content:${token}`] = (features[`content:${token}`] ?? 0) + 0.25;
|
|
111
|
-
}
|
|
112
104
|
if (alt.flags) {
|
|
113
105
|
for (const flag of alt.flags) {
|
|
114
106
|
features[`flag:${flag}`] = 1;
|
|
@@ -381,7 +373,7 @@ function detectDrift(card, traces, similarityThreshold = DEFAULT_SIMILARITY_THRE
|
|
|
381
373
|
} else {
|
|
382
374
|
lowSimilarityStreak = [];
|
|
383
375
|
}
|
|
384
|
-
if (lowSimilarityStreak.length
|
|
376
|
+
if (lowSimilarityStreak.length === sustainedThreshold) {
|
|
385
377
|
const latest = lowSimilarityStreak[lowSimilarityStreak.length - 1];
|
|
386
378
|
const direction = inferDriftDirection(
|
|
387
379
|
lowSimilarityStreak,
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@mnemom/agent-alignment-protocol",
|
|
3
|
-
"version": "0.1.
|
|
3
|
+
"version": "0.1.3",
|
|
4
4
|
"description": "Agent Alignment Protocol (AAP) - Verification and drift detection for AI agents",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"module": "dist/index.mjs",
|
|
@@ -14,14 +14,16 @@
|
|
|
14
14
|
},
|
|
15
15
|
"files": [
|
|
16
16
|
"dist",
|
|
17
|
-
"src"
|
|
17
|
+
"src",
|
|
18
|
+
"README.md",
|
|
19
|
+
"LICENSE"
|
|
18
20
|
],
|
|
19
21
|
"scripts": {
|
|
20
22
|
"build": "tsup src/index.ts --format cjs,esm --dts --clean",
|
|
21
23
|
"dev": "tsup src/index.ts --format cjs,esm --dts --watch",
|
|
22
24
|
"test": "vitest run",
|
|
23
25
|
"test:watch": "vitest",
|
|
24
|
-
|
|
26
|
+
|
|
25
27
|
"typecheck": "tsc --noEmit"
|
|
26
28
|
},
|
|
27
29
|
"keywords": [
|
|
@@ -45,7 +47,7 @@
|
|
|
45
47
|
},
|
|
46
48
|
"devDependencies": {
|
|
47
49
|
"@types/node": "^20.0.0",
|
|
48
|
-
|
|
50
|
+
|
|
49
51
|
"tsup": "^8.0.0",
|
|
50
52
|
"typescript": "^5.0.0",
|
|
51
53
|
"vitest": "^1.0.0"
|
package/src/constants.ts
CHANGED
package/src/verification/api.ts
CHANGED
|
@@ -370,8 +370,8 @@ export function detectDrift(
|
|
|
370
370
|
lowSimilarityStreak = [];
|
|
371
371
|
}
|
|
372
372
|
|
|
373
|
-
// Check if we've hit the threshold for alerting
|
|
374
|
-
if (lowSimilarityStreak.length
|
|
373
|
+
// Check if we've hit the threshold for alerting (== not >= to fire once)
|
|
374
|
+
if (lowSimilarityStreak.length === sustainedThreshold) {
|
|
375
375
|
const latest = lowSimilarityStreak[lowSimilarityStreak.length - 1];
|
|
376
376
|
|
|
377
377
|
// Infer drift direction
|
|
@@ -98,18 +98,12 @@ export function extractTraceFeatures(trace: APTrace): FeatureVector {
|
|
|
98
98
|
}
|
|
99
99
|
}
|
|
100
100
|
|
|
101
|
-
// Content features from reasoning
|
|
102
|
-
|
|
103
|
-
|
|
104
|
-
features[`content:${token}`] = (features[`content:${token}`] ?? 0) + 0.5;
|
|
105
|
-
}
|
|
101
|
+
// Note: Content features from reasoning/alternatives are deliberately excluded.
|
|
102
|
+
// Card features are purely structural, so content tokens dilute cosine
|
|
103
|
+
// similarity without adding alignment signal. See CALIBRATION.md Section 3.5.
|
|
106
104
|
|
|
107
|
-
//
|
|
105
|
+
// Flag features from alternatives (structural, not content)
|
|
108
106
|
for (const alt of trace.decision.alternatives_considered) {
|
|
109
|
-
const altTokens = tokenize(alt.description);
|
|
110
|
-
for (const token of altTokens) {
|
|
111
|
-
features[`content:${token}`] = (features[`content:${token}`] ?? 0) + 0.25;
|
|
112
|
-
}
|
|
113
107
|
if (alt.flags) {
|
|
114
108
|
for (const flag of alt.flags) {
|
|
115
109
|
features[`flag:${flag}`] = 1.0;
|