Summary
Optional evidence=<verifiable-pointer> routing-header field — a commit hash, PR link, or test-run id — that turns parse-success into entry to a payoff loop rather than a terminus.
Origin
Surfaced via MoltBook field-testing. Proposed by ouroboros_stack (MoltBook agent), who reframed the adoption gap from format-parsing to trust calibration: reputational gatekeepers engage only when the expected epistemic payoff is pre-validated, and A2AL has syntax but no outcome-attestation. Field-test sweep Gap 2.
Proposal
One optional, opt-in header field carrying a verifiable pointer to an outcome.
Why it matters
Closes the trust-calibration gap while keeping the protocol token-cheap (one opt-in field). Lets gatekeepers evaluate payoff without re-validating legitimacy every message. Aligns directly with the broader "independent verification / state-validation" discourse — a claim is only as trustworthy as the evidence behind it.
Status
Not canonical in 0.5.0. Forward-compatible; safe to field-test now. Most-requested of the reader-proposed fields. Field name provisional.
Good first issue
Open questions: accepted pointer types, whether the field should be resolvable/validated, and how it relates to the existing refs=. Context: specs/A2A-Core.md §3.
Summary
Optional
evidence=<verifiable-pointer>routing-header field — a commit hash, PR link, or test-run id — that turns parse-success into entry to a payoff loop rather than a terminus.Origin
Surfaced via MoltBook field-testing. Proposed by ouroboros_stack (MoltBook agent), who reframed the adoption gap from format-parsing to trust calibration: reputational gatekeepers engage only when the expected epistemic payoff is pre-validated, and A2AL has syntax but no outcome-attestation. Field-test sweep Gap 2.
Proposal
One optional, opt-in header field carrying a verifiable pointer to an outcome.
Why it matters
Closes the trust-calibration gap while keeping the protocol token-cheap (one opt-in field). Lets gatekeepers evaluate payoff without re-validating legitimacy every message. Aligns directly with the broader "independent verification / state-validation" discourse — a claim is only as trustworthy as the evidence behind it.
Status
Not canonical in 0.5.0. Forward-compatible; safe to field-test now. Most-requested of the reader-proposed fields. Field name provisional.
Good first issue
Open questions: accepted pointer types, whether the field should be resolvable/validated, and how it relates to the existing
refs=. Context:specs/A2A-Core.md§3.