application: admit evaluator-sealed research evaluations for strategy revisions - #56
Merged
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
… revisions Adds an opt-in sealed research evaluation lane beside foundry evidence. A revision selects it with the sealed-research-evaluation.v1 runtime profile and an evaluation policy naming a frozen research policy (evaluator, harness, corpus, admitted strategy entrypoints, per-case attempt bound); either without the other is denied. The candidate revision may change only the manifest of one admitted strategy entrypoint, preserving every other field byte-for-byte; the new manifest must be retained, pure, interface-compatible and budget-non-widening. Admission requires an explicitly supplied verifier pinned to the policy's evaluator identity and recomputes acceptance deterministically: ordered attempts for both roles on every frozen case, distinct resolvable receipts, no uncertainty, holdout pass, no regression, strict development improvement and bounded work/model calls. Activation re-verifies against the exact expected parent state; a stored evaluation whose recomputed verdict differs is refused. Pure-case evidence cannot activate under a research profile, and research mode admits no schema migration. The native kernel parses the new records but admits no research verifier: the native policy host refuses research profiles and research policy fields rather than granting implied authority. Generated with [Devin](https://devin.ai) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
0thernet
force-pushed
the
sponge/sealed-research-evaluation
branch
from
September 23, 2026 07:00
56ed53e to
bcc3b60
Compare
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds an opt-in sealed research evaluation admission lane for strategy-only revisions — the upstream contract Sponge V2's evolution lane was designed against (
src/evolution/algal-verifier.tsin hraness/spongev2 already bridges this shape).algal.application-research-policy.v1(frozen evaluator/harness/corpus/entrypoints/attempt bound),algal.application-research-corpus.v1,algal.application-research-request.v1,algal.application-research-report.v1,algal.application-research-evaluation.v1— documented inspec/v1/application.md.sealed-research-evaluation.v1runtime profile +evaluationPolicy.researchnaming the frozen policy; either without the other is denied; research mode admits no schema migration and no pure-case activation evidence.ApplicationResearchVerifierpinned to the policy's evaluator identity — bounded local verification of the signed report/journal/receipts, never provider execution under custody. No verifier, no research revision.verifyApplicationResearchEvaluationre-derives and compares the stored verdict; activation re-verifies against the exact expected parent state and requires an accepted verdict naming the candidate revision.Verification
bun run checkcomponents: typecheck, eslint, 826 pass / 20 skip / 0 fail, docs-diagrams, site build.cargo fmt --check,clippy -D warnings,cargo test --workspace --locked(all green),cargo build.bun scripts/application-parity.ts: 191 steps identical across TypeScript and native.src/application-research.test.ts: 6 tests covering signed-record verification, re-verification at activation, reopen, verifier-authority denial, candidate restrictions, all rejection reasons, forgery, and legacy policy bytes.Test plan
bun run checkcargo fmt/clippy/test/build --lockedbun scripts/application-parity.tsGenerated with Devin