agentrust-telemetry 0.1.0a3__tar.gz

agentrust_telemetry-0.1.0a3/CHANGELOG.md

@@ -0,0 +1,57 @@
+# Changelog
+
+## Unreleased
+
+## 0.1.0-alpha.3 - 2026-09-07
+
+### Fixed
+
+- The Python `EvidenceAccumulator` deadlocked forever if `append` or `seal` was
+  called reentrantly on the same thread, for example from inside a
+  `durable_append` callback. Both now raise `EvidenceError`, matching the
+  TypeScript SDK's existing protection. `snapshot` remains safely callable
+  reentrantly.
+- Patch four `fast-uri` advisories in the TypeScript lockfile and stop tag
+  interpolation in `release.yml`.
+- `test_trace_adapter_refusals` used a duck-typed signer double, which
+  `agentrust-trace` 0.10.0 rejects. It now builds a real `Ed25519PrivateKey`, so
+  a clean install of this release runs the full suite against either 0.9 or
+  0.10.
+
+### Changed
+
+- Install CI dependencies from hash-pinned lock files, and add `actionlint`
+  plus a test-environment guard to the workflow gates.
+
+## 0.1.0-alpha.2 - 2026-09-02
+
+- Align the wire `spec_version` with `spec/VERSION`; the schema previously
+  pinned a const that appeared nowhere in the packaging.
+- Move the schema `$id` off `agentrust.io`, a domain we do not own, to
+  `agentrust-io.com`.
+- Derive the npm dist-tag from `spec/VERSION` so a prerelease can never publish
+  under `latest`.
+
+## 0.1.0-alpha.1 - 2026-08-19
+
+- Add TypeScript TRACE finalization through a caller-supplied official codec and
+  standardize tool-transcript hashing on RFC 8785 JCS across both SDKs.
+- Add TypeScript metadata-only data-flow primitives plus AGT data-access,
+  audit-policy, and completed-action adapters.
+- Enforce the documented no-URL metadata identifier boundary in both SDKs.
+- Add TypeScript action-bound AGT policy, approval-request, and terminal
+  approval-resolution adapters with deterministic cross-language linkage.
+- Require every AGT approval binding field to be present before comparing it,
+  preventing two absent values from being treated as a valid binding.
+- Add TypeScript OPA, Cedar, and generic AGT policy-decision adapters plus an
+  AGT-compatible fail-closed batch sink.
+- Add TypeScript usage/cost construction, coverage-labelled rollups, and
+  bounded-cardinality OpenTelemetry metric projection.
+- Add a TypeScript evidence accumulator and adopt RFC 8785 JCS for reproducible
+  evidence digests across Python and TypeScript.
+- Encode nanosecond Unix timestamps as canonical decimal strings for exact
+  cross-language JSON behavior.
+- Add the pre-alpha TypeScript reference SDK and shared conformance gates.
+
+- Initial `0.1.0-alpha.1` event contract and conformance fixtures.
+- Initial Python reference SDK with schema/privacy validation and OTel span-event projection.

agentrust_telemetry-0.1.0a3/CODE_OF_CONDUCT.md

@@ -0,0 +1,7 @@
+# Code of Conduct
+
+This project follows the AgentTrust organization Code of Conduct:
+
+https://github.com/agentrust-io/.github/blob/main/CODE_OF_CONDUCT.md
+
+Report conduct concerns through the private contact described there.

agentrust_telemetry-0.1.0a3/CONTRIBUTING.md

@@ -0,0 +1,12 @@
+# Contributing
+
+Contract changes require:
+
+1. a documented semantic reason;
+2. a schema update;
+3. at least one valid fixture;
+4. an invalid or boundary fixture that would catch a plausible defect;
+5. conformance tests;
+6. migration notes for wire-incompatible changes.
+
+No attribute may be added without identifying the signal that carries it and how an instrumentor obtains it at runtime. New content-bearing fields require security and privacy review.

agentrust_telemetry-0.1.0a3/GOVERNANCE.md

@@ -0,0 +1,17 @@
+# Governance
+
+AgentTrust Telemetry is governed under the AgentTrust organization. Maintainers approve releases, normative contract changes, security-sensitive changes, and compatibility claims.
+
+## Decision process
+
+- Routine fixes require maintainer review and passing required checks.
+- Contract changes require an ADR or equivalent design record, valid and invalid fixtures, and conformance coverage.
+- Breaking changes require migration notes and a major contract version change, including during pre-1.0 development when practical.
+- Security and privacy defaults require explicit maintainer approval.
+- Widely applicable telemetry conventions should be proposed upstream to OpenTelemetry rather than permanently duplicated here.
+
+Consensus is preferred. If consensus cannot be reached, maintainers record the decision and dissent in the relevant issue or ADR.
+
+## Releases
+
+Releases are cut from protected `main`, use signed tags where available, and publish provenance and an SBOM. Publishing credentials must use GitHub trusted publishing or another short-lived identity mechanism; long-lived package tokens are prohibited.

agentrust_telemetry-0.1.0a3/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 AgentTrust contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

agentrust_telemetry-0.1.0a3/LIMITATIONS.md

@@ -0,0 +1,25 @@
+# Limitations
+
+Current `0.1.0-alpha.3` limitations:
+
+- The contract and SDK are experimental and may change incompatibly.
+- A Python reference SDK and a TypeScript reference SDK (`packages/typescript/`)
+  exist, with shared-schema and golden-event parity; no other language binding
+  exists yet.
+- OTel span events, Logs, and basic metrics are implemented against caller-owned
+  providers; collector/backend interoperability is not yet exercised.
+- No factory helper emits `approval.cancelled`, the `approval.execution_*`
+  outcomes, or the `evidence.*` lifecycle events; these event types exist in
+  the schema but nothing in either SDK produces them yet.
+- TRACE finalization is software-only and requires explicitly complete evidence
+  plus trusted caller configuration.
+- Action telemetry records resolved attempts only; it does not expose in-flight
+  lifecycle transitions.
+- Evidence memory mode is not durable. Callback mode defines acknowledgement and
+  retry behavior but the adopter owns storage, idempotency, and recovery.
+- Propagation currently supports mutable string mappings; framework-specific HTTP,
+  RPC, and messaging carrier adapters are not yet included.
+- The SDK validates declared metadata but cannot prove a producer's policy decision, identity, classification, token count, or cost is truthful.
+- Operational OTel delivery may be sampled or dropped and is not durable audit evidence.
+- The current schemas do not provide a general content-capture profile.
+- Interoperability has been exercised locally with OpenTelemetry Python, not across collectors, backends, or languages.

agentrust_telemetry-0.1.0a3/MAINTAINERS.md

@@ -0,0 +1,7 @@
+# Maintainers
+
+Initial maintainer:
+
+- Imran Siddique — GitHub: `imran-siddique`
+
+Maintainers are responsible for contract integrity, privacy and security review, releases, and contributor conduct. Changes to this list require an approved pull request.

agentrust_telemetry-0.1.0a3/MANIFEST.in

@@ -0,0 +1,20 @@
+include CHANGELOG.md
+include CODE_OF_CONDUCT.md
+include CONTRIBUTING.md
+include GOVERNANCE.md
+include LIMITATIONS.md
+include MAINTAINERS.md
+include NOTICE
+include PRIVACY.md
+include RELEASING.md
+include REPOSITORY-SETUP.md
+include ROADMAP.md
+include SECURITY.md
+include SUPPORT.md
+include VERSIONING.md
+recursive-include compatibility *.json
+recursive-include conformance *.json *.py *.txt
+recursive-include docs *.md
+recursive-include examples *.py
+recursive-include spec VERSION *.json
+recursive-include tools *.py

agentrust_telemetry-0.1.0a3/NOTICE

@@ -0,0 +1,4 @@
+AgentTrust Telemetry
+Copyright 2026 AgentTrust contributors
+
+This product includes concepts and terminology derived from OpenTelemetry semantic conventions. OpenTelemetry is a project of the Cloud Native Computing Foundation.

agentrust_telemetry-0.1.0a3/PKG-INFO

@@ -0,0 +1,180 @@
+Metadata-Version: 2.4
+Name: agentrust-telemetry
+Version: 0.1.0a3
+Summary: Backend-neutral governance telemetry for AI-agent runtimes
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/agentrust-io/agentrust-telemetry
+Project-URL: Repository, https://github.com/agentrust-io/agentrust-telemetry
+Project-URL: Issues, https://github.com/agentrust-io/agentrust-telemetry/issues
+Project-URL: Security, https://github.com/agentrust-io/agentrust-telemetry/security/policy
+Keywords: ai-agents,governance,opentelemetry,observability,telemetry
+Classifier: Development Status :: 3 - Alpha
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Typing :: Typed
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+License-File: NOTICE
+Requires-Dist: jsonschema[format]<5,>=4.23
+Requires-Dist: referencing<1,>=0.35
+Requires-Dist: rfc8785<1,>=0.1.4
+Provides-Extra: otel
+Requires-Dist: opentelemetry-api<2,>=1.30; extra == "otel"
+Provides-Extra: trace
+Requires-Dist: agentrust-trace<1,>=0.9; python_version >= "3.11" and extra == "trace"
+Provides-Extra: test
+Requires-Dist: opentelemetry-sdk<2,>=1.30; extra == "test"
+Requires-Dist: agentrust-trace<1,>=0.9; python_version >= "3.11" and extra == "test"
+Requires-Dist: tomli<3,>=2; python_version < "3.11" and extra == "test"
+Dynamic: license-file
+
+# AgentTrust Telemetry
+
+Portable governance telemetry and verifiable evidence for AI-agent runtimes.
+
+This repository defines a backend-neutral contract for policy decisions, approval lifecycles, usage, classified data flows, and evidence lifecycle events. It composes with OpenTelemetry; it is not a tracing backend, policy engine, agent framework, or dashboard.
+
+> **Status:** alpha contract `0.1.0-alpha.3`. No stable SDK API or compatibility guarantee exists yet.
+
+## Why
+
+Agent applications can already emit model and tool traces, but governance facts are often trapped in policy-engine logs, approval databases, cost modules, and proprietary dashboards. AgentTrust Telemetry gives those facts one privacy-conscious contract and correlates them with the OpenTelemetry trace already produced by the application.
+
+The application keeps its collector, backend, policy engine, workflow framework, and UI.
+
+## Event families
+
+| Family | Purpose |
+|---|---|
+| Policy decision | Allow, deny, challenge, error, enforcement mode, policy identity and timing |
+| Approval lifecycle | Requested through terminal decision and execution outcome, bound to an action digest |
+| Usage | Per-call/run token and cost facts with explicit cost provenance |
+| Data flow | Classified source-to-destination metadata without payload capture |
+| Action execution | Resolved tool, MCP, A2A, file, HTTP, and database attempts |
+| Evidence lifecycle | Run checkpoints, completeness, and optional TRACE finalization status |
+
+## Current contents
+
+- JSON Schema 2020-12 event contracts in `spec/schema/`.
+- Valid and invalid conformance fixtures in `conformance/fixtures/`.
+- An independent Python conformance runner in `conformance/runner/`.
+- Contract tests in `tests/`.
+
+## Validate fixtures
+
+```shell
+python -m pip install -r conformance/requirements.txt
+python conformance/runner/validate.py
+python -m unittest discover -s tests -v
+```
+
+The validator checks schema conformance and the metadata-only privacy invariant. It does not yet validate OTLP projection or TRACE mapping.
+
+## Python reference SDK
+
+Install from a checkout while the project is pre-release:
+
+```shell
+python -m pip install -e ".[otel]"
+```
+
+```python
+from agentrust_telemetry import SchemaValidator, TelemetryClient
+
+client = TelemetryClient(SchemaValidator.bundled())
+result = client.emit(normalized_event)
+```
+
+The SDK uses the caller's current OpenTelemetry span when `opentelemetry-api` is installed. It never installs a provider or exporter. A caller may additionally supply a structured-log emitter.
+
+For synchronous cross-process agent calls, propagate the caller's W3C context and
+durable AgenTrust identifiers, then use the extracted context as the receiving
+span's remote parent:
+
+```python
+from agentrust_telemetry import extract_context, inject_context
+
+carrier = {}
+inject_context(carrier, run_id="run-123", agent_id="planner")
+
+remote = extract_context(carrier)
+with tracer.start_as_current_span("worker", context=remote.otel_context):
+    event.update(remote.event_fields(agent_id="worker"))
+```
+
+For asynchronous queue handoffs, start a new trace with
+`links=[remote.link()]`. This preserves causality without representing queued
+work as a synchronous child span. Propagated metadata is untrusted input and
+does not establish agent identity or authorization.
+
+Run the synthetic example:
+
+```shell
+python examples/manual_governance.py
+```
+
+## TypeScript reference SDK
+
+The pre-alpha Node package lives in `packages/typescript`:
+
+```shell
+cd packages/typescript
+npm ci
+npm run check
+```
+
+It validates the same fixtures and privacy profile as Python and preserves
+nanosecond wire timestamps as decimal strings. It does not install an OTel
+provider, exporter, or global propagator.
+
+Run the complete AGT-compatible governance, OTel, durable-evidence, and TRACE
+reference workflow (Python 3.11+ with test extras installed):
+
+```shell
+python -m pip install -e ".[test]"
+python examples/governed_workflow.py
+```
+
+## Contract principles
+
+- `run_id` is durable execution correlation; `trace_id` is an optional W3C operational trace context.
+- Standard OpenTelemetry fields take precedence over AgentTrust extensions.
+- Raw prompts, output, source code, tool arguments/results, credentials, and authorization tokens are prohibited in the metadata-only profile.
+- Operational telemetry can be lossy; evidence completeness must never be overstated.
+- An event reports a fact observed elsewhere. This project does not make authorization decisions.
+
+## Architecture and project policy
+
+- [Architecture](docs/architecture.md)
+- [OpenTelemetry projection](docs/otel-projection.md)
+- [OpenTelemetry GenAI compatibility](docs/otel-genai-compatibility.md)
+- [Action execution events](docs/action-events.md)
+- [Data-flow classification](docs/data-flow-classification.md)
+- [Usage and cost attribution](docs/usage-attribution.md)
+- [Event factories and policy adapters](docs/adapters.md)
+- [Evidence chain profile](docs/evidence-chain.md)
+- [TRACE finalization](docs/trace-finalization.md)
+- [Privacy](PRIVACY.md)
+- [Limitations](LIMITATIONS.md)
+- [Roadmap](ROADMAP.md)
+- [Security](SECURITY.md)
+- [Governance](GOVERNANCE.md)
+- [Sponsors](SPONSORS.md)
+- [Releasing](RELEASING.md)
+- [Contributing](CONTRIBUTING.md)
+
+## What this project does not provide
+
+- A telemetry collector, storage service, dashboard, or SaaS backend.
+- Agent/model auto-instrumentation that duplicates OpenTelemetry GenAI or OpenInference.
+- Policy evaluation or human-approval workflow execution.
+- A model pricing catalog.
+- A claim that sampled operational telemetry is durable audit evidence.
+
+## License
+
+MIT. See `LICENSE`.

agentrust_telemetry-0.1.0a3/PRIVACY.md

@@ -0,0 +1,17 @@
+# Privacy
+
+## Default profile
+
+`metadata_only` is the mandatory default. The SDK rejects known content-bearing fields and rejects extension attributes unless the application explicitly allowlists each key.
+
+The default prohibits prompts, completions, source code, tool arguments/results, credentials, secrets, and authorization tokens. Content digests are identifiers and may still be sensitive; operators must treat them according to their threat model.
+
+## Responsibilities
+
+Instrumentation authors must supply pseudonymous identities, accurate classification labels, and minimal metadata. Operators remain responsible for collector transport security, access control, retention, deletion, residency, and compliance obligations.
+
+The SDK records classification supplied by another component. It does not establish that the classification is correct.
+
+## Opt-in extensions
+
+Extension attributes require an explicit allowlist in `SchemaValidator`. Allowlisting a key is a deployment decision, not proof that every value under that key is safe. Broad wildcards are intentionally unsupported.

agentrust_telemetry-0.1.0a3/README.md

@@ -0,0 +1,146 @@
+# AgentTrust Telemetry
+
+Portable governance telemetry and verifiable evidence for AI-agent runtimes.
+
+This repository defines a backend-neutral contract for policy decisions, approval lifecycles, usage, classified data flows, and evidence lifecycle events. It composes with OpenTelemetry; it is not a tracing backend, policy engine, agent framework, or dashboard.
+
+> **Status:** alpha contract `0.1.0-alpha.3`. No stable SDK API or compatibility guarantee exists yet.
+
+## Why
+
+Agent applications can already emit model and tool traces, but governance facts are often trapped in policy-engine logs, approval databases, cost modules, and proprietary dashboards. AgentTrust Telemetry gives those facts one privacy-conscious contract and correlates them with the OpenTelemetry trace already produced by the application.
+
+The application keeps its collector, backend, policy engine, workflow framework, and UI.
+
+## Event families
+
+| Family | Purpose |
+|---|---|
+| Policy decision | Allow, deny, challenge, error, enforcement mode, policy identity and timing |
+| Approval lifecycle | Requested through terminal decision and execution outcome, bound to an action digest |
+| Usage | Per-call/run token and cost facts with explicit cost provenance |
+| Data flow | Classified source-to-destination metadata without payload capture |
+| Action execution | Resolved tool, MCP, A2A, file, HTTP, and database attempts |
+| Evidence lifecycle | Run checkpoints, completeness, and optional TRACE finalization status |
+
+## Current contents
+
+- JSON Schema 2020-12 event contracts in `spec/schema/`.
+- Valid and invalid conformance fixtures in `conformance/fixtures/`.
+- An independent Python conformance runner in `conformance/runner/`.
+- Contract tests in `tests/`.
+
+## Validate fixtures
+
+```shell
+python -m pip install -r conformance/requirements.txt
+python conformance/runner/validate.py
+python -m unittest discover -s tests -v
+```
+
+The validator checks schema conformance and the metadata-only privacy invariant. It does not yet validate OTLP projection or TRACE mapping.
+
+## Python reference SDK
+
+Install from a checkout while the project is pre-release:
+
+```shell
+python -m pip install -e ".[otel]"
+```
+
+```python
+from agentrust_telemetry import SchemaValidator, TelemetryClient
+
+client = TelemetryClient(SchemaValidator.bundled())
+result = client.emit(normalized_event)
+```
+
+The SDK uses the caller's current OpenTelemetry span when `opentelemetry-api` is installed. It never installs a provider or exporter. A caller may additionally supply a structured-log emitter.
+
+For synchronous cross-process agent calls, propagate the caller's W3C context and
+durable AgenTrust identifiers, then use the extracted context as the receiving
+span's remote parent:
+
+```python
+from agentrust_telemetry import extract_context, inject_context
+
+carrier = {}
+inject_context(carrier, run_id="run-123", agent_id="planner")
+
+remote = extract_context(carrier)
+with tracer.start_as_current_span("worker", context=remote.otel_context):
+    event.update(remote.event_fields(agent_id="worker"))
+```
+
+For asynchronous queue handoffs, start a new trace with
+`links=[remote.link()]`. This preserves causality without representing queued
+work as a synchronous child span. Propagated metadata is untrusted input and
+does not establish agent identity or authorization.
+
+Run the synthetic example:
+
+```shell
+python examples/manual_governance.py
+```
+
+## TypeScript reference SDK
+
+The pre-alpha Node package lives in `packages/typescript`:
+
+```shell
+cd packages/typescript
+npm ci
+npm run check
+```
+
+It validates the same fixtures and privacy profile as Python and preserves
+nanosecond wire timestamps as decimal strings. It does not install an OTel
+provider, exporter, or global propagator.
+
+Run the complete AGT-compatible governance, OTel, durable-evidence, and TRACE
+reference workflow (Python 3.11+ with test extras installed):
+
+```shell
+python -m pip install -e ".[test]"
+python examples/governed_workflow.py
+```
+
+## Contract principles
+
+- `run_id` is durable execution correlation; `trace_id` is an optional W3C operational trace context.
+- Standard OpenTelemetry fields take precedence over AgentTrust extensions.
+- Raw prompts, output, source code, tool arguments/results, credentials, and authorization tokens are prohibited in the metadata-only profile.
+- Operational telemetry can be lossy; evidence completeness must never be overstated.
+- An event reports a fact observed elsewhere. This project does not make authorization decisions.
+
+## Architecture and project policy
+
+- [Architecture](docs/architecture.md)
+- [OpenTelemetry projection](docs/otel-projection.md)
+- [OpenTelemetry GenAI compatibility](docs/otel-genai-compatibility.md)
+- [Action execution events](docs/action-events.md)
+- [Data-flow classification](docs/data-flow-classification.md)
+- [Usage and cost attribution](docs/usage-attribution.md)
+- [Event factories and policy adapters](docs/adapters.md)
+- [Evidence chain profile](docs/evidence-chain.md)
+- [TRACE finalization](docs/trace-finalization.md)
+- [Privacy](PRIVACY.md)
+- [Limitations](LIMITATIONS.md)
+- [Roadmap](ROADMAP.md)
+- [Security](SECURITY.md)
+- [Governance](GOVERNANCE.md)
+- [Sponsors](SPONSORS.md)
+- [Releasing](RELEASING.md)
+- [Contributing](CONTRIBUTING.md)
+
+## What this project does not provide
+
+- A telemetry collector, storage service, dashboard, or SaaS backend.
+- Agent/model auto-instrumentation that duplicates OpenTelemetry GenAI or OpenInference.
+- Policy evaluation or human-approval workflow execution.
+- A model pricing catalog.
+- A claim that sampled operational telemetry is durable audit evidence.
+
+## License
+
+MIT. See `LICENSE`.

agentrust_telemetry-0.1.0a3/RELEASING.md

@@ -0,0 +1,59 @@
+# Releasing
+
+Releases publish the Python and TypeScript SDKs from one GitHub release event.
+Do not build or upload either package from a maintainer workstation.
+
+## One-time registry setup
+
+Create protected GitHub environments named `pypi` and `npm`, each with a
+required maintainer review and deployment restricted to protected tags.
+
+Configure PyPI pending trusted publishing with:
+
+- project: `agentrust-telemetry`
+- owner/repository: `agentrust-io/agentrust-telemetry`
+- workflow: `release.yml`
+- environment: `pypi`
+
+Configure the npm trusted publisher for `@agentrust-io/telemetry` with:
+
+- organization/repository: `agentrust-io/agentrust-telemetry`
+- workflow: `release.yml`
+- environment: `npm`
+- allowed action: `npm publish`
+
+The npm package must be owned by the `agentrust-io` npm organization before its
+trusted publisher can be configured. If npm does not expose publisher settings
+until the first version exists, bootstrap only that first package ownership
+using npm's interactive 2FA flow, then configure trusted publishing before any
+subsequent release. Never store an npm publish token in GitHub.
+
+## Release procedure
+
+1. Move the changelog entries from `Unreleased` into the new dated version.
+2. Update `spec/VERSION`; the Python and npm spellings are checked by
+   `tools/check_versions.py`.
+3. Run the complete CI and artifact smoke tests through a pull request.
+4. Merge the release-preparation pull request to `main`.
+5. Create a GitHub prerelease or release targeting `main`, with tag
+   `v<contract-version>`.
+6. Approve the `pypi` and `npm` deployment jobs after inspecting their exact
+   source commit and built artifacts.
+7. Confirm both registry versions, GitHub release assets, and provenance before
+   announcing the release.
+
+The workflow fails closed if the tag differs from the contract version. It
+builds each distribution once, sends the same artifacts to the registries, and
+attaches them to the GitHub release.
+
+The npm dist-tag is derived from `spec/VERSION` by `tools/npm_dist_tag.py`, so a
+prerelease publishes under `alpha`, `beta`, `rc` or `dev` and only a stable
+version publishes under `latest`. npm refuses an untagged prerelease publish
+outright, and an untagged stable publish would move `latest`, so nothing here is
+left to the person running the release. If you ever must publish by hand, pass
+the same tag: `npm publish --tag "$(python tools/npm_dist_tag.py)"`.
+
+Note that npm sets `latest` on a package's very first published version whatever
+`--tag` says. That is expected on a bootstrap publish and corrects itself when
+the first stable version ships. PyPI and npm create registry provenance
+through trusted publishing; GitHub also attests the downloadable release assets.

agentrust_telemetry-0.1.0a3/REPOSITORY-SETUP.md

@@ -0,0 +1,42 @@
+# GitHub repository creation checklist
+
+Verified on 2026-08-18: `agentrust-io/agentrust-telemetry` does not exist and the name is available.
+
+PyPI also returned no matching distribution for `agentrust-telemetry` on 2026-08-18. Availability is not reservation; re-check immediately before the first publish.
+
+## Creation parameters
+
+- Owner: `agentrust-io`
+- Name: `agentrust-telemetry`
+- Visibility: public
+- Default branch: `main`
+- Description: `Backend-neutral governance telemetry and verifiable evidence for AI-agent runtimes.`
+- Topics: `ai-agents`, `governance`, `opentelemetry`, `observability`, `telemetry`, `trace`
+- Initialize remotely: no README, license, or gitignore; this repository already supplies them.
+
+## Settings immediately after first push
+
+- Enable private vulnerability reporting.
+- Enable secret scanning and push protection where available.
+- Enable Dependabot security updates.
+- Enable automatic branch deletion after merge.
+- Disable merge commits; allow squash merge with PR title as the default message.
+- Protect `main` with pull requests, one approving review, CODEOWNER review, dismissal of stale approvals, conversation resolution, and required status checks.
+- Required checks after their first run: all CI matrix jobs, CodeQL, and OpenSSF Scorecard where GitHub permits it.
+- Prevent force pushes and branch deletion.
+- Require signed commits if that matches organization-wide practice; do not enable until maintainers can comply.
+- Configure the `pypi` environment and trusted publishing only when the package name and first release are approved.
+
+## First push sequence
+
+1. Review the complete staged diff and generated schema copies.
+2. Run `python tools/check_versions.py`.
+3. Run `python tools/check_schemas.py`.
+4. Run `python conformance/runner/validate.py`.
+5. Run `python -m unittest discover -s tests -v`.
+6. Run `python -m build` and `python tools/smoke_wheel.py <wheel>`.
+7. Create one DCO-signed initial commit.
+8. Create the empty GitHub repository with the parameters above.
+9. Add `origin`, push `main`, then apply the settings above.
+
+Repository creation, initial commit, push, package-name reservation, and publishing are intentionally not performed by readiness work.

agentrust_telemetry-0.1.0a3/ROADMAP.md

@@ -0,0 +1,24 @@
+# Roadmap
+
+## v0.1 contract and Python reference
+
+- Normalized event schemas and conformance fixtures.
+- Metadata-only privacy enforcement.
+- Caller-owned OTel span-event projection.
+- Structured-log projection contract.
+- Reproducible package and CI gates.
+- Per-call usage facts and coverage-labelled per-agent/workflow cost rollups.
+- Executable OpenTelemetry GenAI compatibility matrix pinned to upstream source.
+- TypeScript reference primitives with shared-schema and golden-event parity.
+
+## Next
+
+- Additional metric views and collector interoperability fixtures.
+- Additional end-to-end scenarios for failure, asynchronous handoff, and incomplete evidence.
+- Additional classifier adapters and organization taxonomy profiles.
+- Additional W3C propagation carrier adapters beyond mutable string mappings.
+- Action lifecycle expansion if adopters need in-flight attempt telemetry.
+- Framework integrations and future consolidated-core native event surfaces.
+- Additional TypeScript adapters and .NET, Rust, and Go reference primitives.
+
+Roadmap items are intentions, not shipped behavior or compatibility commitments.