Official descriptionOpen source7d ago
probityai/agent-evidence-observer
Records brokered file requests from AI agents and includes an offline reader for selected native delegation…
View on GitHub ↗Direct link · no redirect page
- Star count
- 2 ★ · probityai/agent-evidence-observer · read from the GitHub search API · (checked 2026-10-16 20:00:20)
- Source
- read directly from its GitHub page · type: open-source project
- Harvest history
- first seen 2026-10-10 · last seen 2026-10-16 · seen 13× · listed by 1 sources
- Score
- score 103 · 2 ★ · 1 sources · repo
FAQ (answers come from our records)
How do I install it?
We did not find an install command for this repository (our install coverage is only about 4%). Check its README or releases page.
Who is it for?
We filed it under these tags: 开源项目. The tags come from the entry itself, not a taxonomy we imposed.
Who maintains it?
repository
probityai/agent-evidence-observer；first catalogued 2026-10-10；last seen 2026-10-16。There is no official certification or quality rating here — we record sources and dates only.Third-party quality signals
Agent Evidence Observer (prototype)
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
Record brokered agent writes and check selected native framework reports offline.
Use it in agent harnesses and evaluation pipelines that need replayable action records or a consumer gate for retained reports.
Quick start
These commands install a pinned source snapshot. Python 3.12 or later and
uv are needed:
The two commands print:
sample-run/ now holds the signed packet, the history, the public keys and the file tree. The demo
writes a file, retries it, and the retry returns the first effect. noDetectedGap means the broker's
file-tree snapshots found no change it did not make. It does not mean every agent effect was seen.
Status
The root distribution is version 0.0.1. Every record says witnessScope: PEER: the witness key runs on the same
host under the same operator, so the records show what this broker saw, not what an independent
observer saw. Linux runs can launch a fixed probe, or a Python script given to run-workload, under
bubblewrap with only the broker's socket exposed. Running a real agent that way, and a separately operated
witness, are the next steps.
Documentation
Start with worked examples to follow a practical question,
then open its technical reference for commands and exact evidence.
| page | read it for |
| --- | --- |
| Component tour | every part of the prototype in one paragraph each, with links: ticket service, replay, evaluation history, A2A, AAE, protected dispatch |
| Running the demo | the full install with tests, and what each output field means |
| Design, trust boundary and roadmap | what each mechanism checks and does not establish, the Linux boundary gate, recovery, consumer admission, and the roadmap |
| Isolated producer | the acceptance criteria for running an unmodified agent behind the broker |
| Brokered workload | run a Python script behind the write broker with run-workload, then check it offline with verify-workload |
| Witness ledger | signed begin and terminal receipts, and how a reader checks them |
| Installation and profile selection | choose the root wheel, a separate profile wheel or a selected source reader; runtime requirements and installed examples |
| Release the root package | qualify a wheel and source distribution, then publish with a bound source tag |
| Atomic native reader | install an offline reader for selected native delegation runs and their refusal controls |
Use it in agent harnesses and evaluation pipelines that need replayable action records or a consumer gate for retained reports.
Quick start
These commands install a pinned source snapshot. Python 3.12 or later and
uv are needed:
The two commands print:
sample-run/ now holds the signed packet, the history, the public keys and the file tree. The demo
writes a file, retries it, and the retry returns the first effect. noDetectedGap means the broker's
file-tree snapshots found no change it did not make. It does not mean every agent effect was seen.
Status
The root distribution is version 0.0.1. Every record says witnessScope: PEER: the witness key runs on the same
host under the same operator, so the records show what this broker saw, not what an independent
observer saw. Linux runs can launch a fixed probe, or a Python script given to run-workload, under
bubblewrap with only the broker's socket exposed. Running a real agent that way, and a separately operated
witness, are the next steps.
Documentation
Start with worked examples to follow a practical question,
then open its technical reference for commands and exact evidence.
| page | read it for |
| --- | --- |
| Component tour | every part of the prototype in one paragraph each, with links: ticket service, replay, evaluation history, A2A, AAE, protected dispatch |
| Running the demo | the full install with tests, and what each output field means |
| Design, trust boundary and roadmap | what each mechanism checks and does not establish, the Linux boundary gate, recovery, consumer admission, and the roadmap |
| Isolated producer | the acceptance criteria for running an unmodified agent behind the broker |
| Brokered workload | run a Python script behind the write broker with run-workload, then check it offline with verify-workload |
| Witness ledger | signed begin and terminal receipts, and how a reader checks them |
| Installation and profile selection | choose the root wheel, a separate profile wheel or a selected source reader; runtime requirements and installed examples |
| Release the root package | qualify a wheel and source distribution, then publish with a bound source tag |
| Atomic native reader | install an offline reader for selected native delegation runs and their refusal controls |
- Source
- github · https://github.com/probityai/agent-evidence-observer
- Browse by field
- Open source
- Traction
- 2 ★
- added
- yesterday
More in Open source
Filed under
Every tag opens the full list for that filter. Counts are the live catalogue figures.