Inputs
- Target reference and action record
- Observation configuration
- Effect store or adapter exposure
Runtime and execution assurance
Independent Observer
A response is not proof of what happened.
Post-action observation of external state feeding reconciliation — with the current independence boundary stated explicitly.
Responsibility
What Independent Observer is responsible for
Independent Observer captures external state after an action so reconciliation compares observed reality with the required postcondition instead of trusting the executor's response. In the reviewed implementation the read-back is performed by NARG against the same local effect store; a separate third-party reader has not been built.
An executor saying "success" is a claim. Observation is the evidence that something actually changed. The Observer's job is to capture the post-action state that reconciliation needs, including the case where state cannot be observed.
The honest boundary is independence. The reviewed implementation performs a real read-back, but through NARG and the same effect store. A marker adapter cannot detect external drift. The site therefore describes the capability as demonstrable, not as an independent third-party observer.
Lifecycle position
Where it sits in the governance lifecycle
Stage numbering is canonical across the site. Open the interactive lifecycle to see every component with responsibilities at each stage.
Independent observation
Record what the external state actually is.
Verified capabilities
What is verified today
Capabilities are drawn from the product register. Each one is bounded by the limitations stated further down this page.
- Post-action state capture and read-back
- Observed state supplied to reconciliation
- Explicit unknown observation when state cannot be read
- Divergence detection when observed state contradicts the claimed outcome
Outputs
- Observed state record
- Observation result (observed / unknown)
- Input to reconciliation
Evidence and maturity
What the evidence supports
Maturity describes verified engineering evidence in a bounded environment. It is not a production-readiness statement.
Maturity note
Observation and read-back were exercised on the recorded chain, and divergent and unknown outcomes were detected. The observer is not yet a third-party process: it reads the same local effect store the dispatcher writes.
Not production-qualified
This label describes verified engineering evidence only. No production deployment, certification, benchmark or customer outcome is claimed.
Evidence recorded
- Recorded chain hop 10: execute plus effect read-back.
- Conformance matrix structural gap 2: the effect is observed by NARG, not a separate reader.
- Acceptance step 14: mismatched observation produced DIVERGED; unobservable produced UNKNOWN.
Dependencies
- NARG
- Substrate
- Reconciliation
Dependencies are stated as the product register records them. A dependency is not automatically production-qualified; see its own product or component page for its maturity and limitations.
Limitations
What this product does not do (yet).
Limitations are stated before the evaluation questions so they cannot be missed, downplayed or treated as future-work footnotes.
Limitations — read before evaluating
- Not a third-party independent reader: observation is performed by NARG against the same local effect store.
- The marker adapter cannot detect external drift.
- No separate observer process or remote attestation of observed state is implemented.
- Independence is a design target, not a current guarantee.
No capability on this page should be read as production qualification. If a limitation is material to your use case, it belongs in the evaluation scope — not after it.
Relationship to the platform
How it fits
The Observer supplies the factual input to reconciliation. It does not authorise actions and does not decide conformance; it records what the external state is, including when it cannot be established.
Evaluation questions
- What counts as observed state, and can it be produced without the executor's cooperation?
- How is an unobservable outcome represented?
- What would third-party independence require in your environment?
Next step
Evaluate Independent Observer against a real control problem.
Bring a system boundary, an action class and a failure mode. The first step is to agree what should be governed and what evidence would count.