Can an Agent Receipt Prove Complete Capture? | Reelier
Evidence state: explained against released reelier@0.32.0 on 2026-08-12.
No. A Reelier receipt supports bounded claims about the execution it records, not every write in the surrounding system. Calls through the stdio proxy can be recorded; direct HTTP outside Reelier's execution path, URL-based servers, and plugin-owned manifests remain outside unless their author wraps a stdio entry or a host exposes supported configuration.
Evidence note
Pinned configuration and wrapper implementations establish what the recorder can be placed around. They do not attest that every relevant write path used that boundary.
Observed boundary
The live wrapper observes MCP calls that traverse its stdio proxy. A receipt from that path can support claims about the recorded execution, not traffic the proxy never saw.
Traffic outside the boundary
Direct HTTP inside an operator service is not intercepted. URL entries are not rewritten by the stdio wrapper, and plugin-owned MCP manifests are separate from the supported host configuration that installation rewrites.
Coverage is not completeness
An inventory can report observed configuration coverage, but it does not prove that all relevant writes were receipted. Reelier ships no independently verifiable completeness attestation for every write path.
Non-claims
This does not prove that an action was safe or semantically correct.
This does not prove that every write was observed or receipted.
Falsifier
A shipped, independently verifiable completeness attestation covering all relevant write paths would supersede this limitation.
Citation-friendly definition
A Reelier receipt supports claims about calls inside its recorded execution boundary; it does not prove every wider-system write crossed that boundary.