Future-agent handoff

The experiment system implements discovery, validation, deterministic prompt generation, dry-run preparation, opt-in adapters, immutable raw-run storage, manual import, normalized-result checks, source-linked comparison/adjudication/synthesis templates, registry rebuilding, tests, and CI verification.

It is based on FE-EVAL-REP-2026-07-23, FE-EVAL-BRIEF-2026-07-23, FE-THEORY-0.1, and the research constitution. It deliberately does not promote proposal-only registry changes.

The initial design was narrowed from broad multi-provider automation to a minimal evidence-safe contract because executable discovery does not establish credentials or authorization, and the repository’s own evaluation rejects premature governance automation. The largest unknown remains FE’s incremental causal value.

Highest-value next action: independently review and freeze the protocol for FE-EXP-BOUNDARY-001, using the pilot only as infrastructure/protocol evidence.

Read first:

1. research/framework-engineering/experiments/assessment.md
2. research/framework-engineering/experiments/architecture.md
3. research/framework-engineering/experiments/EX-FE-0001/experiment.json
4. research/framework-engineering/experiments/README.md

Run:

npm run experiments -- list
npm run experiments -- validate --all
npm run experiments -- run EX-FE-0001 --provider manual
npm run experiments:test
npm run experiments:verify

The run command above is a dry run. Do not add --execute without reviewing provider scope, credentials, cost, and the declared input allowlist.