agentfootprint
Version:
The explainable agent framework — backtrack a wrong answer to the exact context that caused it (evidence, not guesses). Built on footprintjs.
18 lines • 954 B
JavaScript
/**
* scriptRecord — the vocabulary of a code call on the record: the label its
* execution record earns, the verdict on a fenced block an answer shows, and
* the attachment that travels with an answer.
*
* Pattern: types only — the words two layers must agree on and cannot import
* each other for: the agent's script record (`core/agent/scripts/`,
* which re-exports every name here) and the event payloads that carry
* it (`events/payloads.ts` · `AgentEvidenceCheckedPayload`, `AgentTurnEndPayload`).
* Role: lib/ leaf (L0). No runtime code.
*
* Design: docs/design/2026-10-script-artifact.md. FOUR FACTS, KEPT APART:
* "this is the script submitted" and "this script ran" are facts about a CALL;
* "a tool returned this value" is the evidence gate's, from `role: 'tool'`
* results only; "this claim holds" has no record. No type here joins them.
*/
export {};
//# sourceMappingURL=scriptRecord.js.map