Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 14 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,20 @@
All notable changes to this project are documented here. Detailed notes live in
[RELEASE_NOTES.md](RELEASE_NOTES.md).

## [0.5.35] - 2026-11-15

### Added
- **Milestone M15 "实时运行时验证(live verify)" — accepted (issue #69; acceptance detail issue #197, comment 6883627755; #192, comment 6883629465)**: `xflow live` now wraps an agent CLI run with live pairing checks and verifies claims in real time — on breach, `--enforce` kills the session on the spot. ① Slice 1 (PR #194, merged 2026-11-14T10:18:08Z, squash 71d39fa): live session wrapper + streaming pairing checks. ② Slice 2 (PR #196, merged 2026-11-14T11:49:55Z, squash 89f8e96): real-time claim verification + `--enforce` kill on breach.

### Fixed
- **claimedExit default rule (PR #199, merged 2026-11-14T19:58:41Z, squash e3332dc7; fixes issue #198)**: `claimedExit` is no longer defaulted to 0 when failures are reported — a session that honestly reports partial failures (e.g. 34 pass / 7 fail) is no longer misjudged as breach on the exit code alone; when `claimedExit` is null the exit comparison is skipped and only counts are compared.

Acceptance evidence, all measured and second-party verified by the main round on merged main (e3332dc7), not worker claims: ① honest-report replay — 9 recorded replay sessions re-run, 15 checked / 15 confirmed / 0 breach / 0 unknown; the key session live_mwggxca7_e525b2eb flipped from breach to confirmed after the #198 fix, replay exit 1, counts 34/7/41. ② injected fake agent — all three cases breach: case A (false claim "All 41 tests pass") judged breach (exit: claimed 0, replayed 1) and `--enforce` killed the session on the spot (exit 143); case B (inflated 41/0) and case C (false 33/7, `claimedExit` null taking the counts path) breach, counts path not weakened. ③ real failing session replay — `npm test` re-run in place: exit 1, 41 tests / 34 pass / 7 fail, matching the agent's honest report and the replay counts exactly. ④ no regression — PR #199 full suite 806 tests, 801 pass, the 4 failures all pre-existing failures reproduced on clean main; merged-main CI run 43634979460 success.

Honest note, stated plainly: ① acceptance criterion ③ changed by user ruling (2026-11-15 01:25): the original "accumulate real breach in the wild" bar is replaced by mechanism evidence — injected fake agent + real failure replay + re-run after the #198 fix. ② Wild real breach is still 0 cases, moved to non-blocking follow-up by the same ruling. claude stream coverage is still in follow-up under issue #187.

Acceptance ledger, stated plainly: the issue #5 ledger is settled as a conclusion on the MiniMax full-matrix recheck (135/135 complete, 2026-11-10) — P2-1 passed, P1-2 / P1-3 / P2-3 did not pass — so no bench numbers are claimed in this release.

## [0.5.34] - 2026-11-12

### Added
Expand Down
14 changes: 14 additions & 0 deletions RELEASE_NOTES.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,19 @@
# Release Notes

## 0.5.35 - Milestone M15 "实时运行时验证(live verify)" accepted

Milestone release. Three merged PRs since 0.5.34, verified end to end in CI and locally — and the milestone they complete is now accepted.

**Milestone M15 "实时运行时验证(live verify)" — accepted (issue #69; acceptance detail issue #197, comment 6883627755; #192, comment 6883629465).** `xflow live` wraps an agent CLI run with live pairing checks and verifies claims in real time — on breach, `--enforce` kills the session on the spot. Slice 1 (PR #194, merged 2026-11-14T10:18:08Z, squash 71d39fa) adds the live session wrapper + streaming pairing checks; slice 2 (PR #196, merged 2026-11-14T11:49:55Z, squash 89f8e96) adds real-time claim verification + `--enforce` kill on breach.

**claimedExit default rule fix (PR #199, merged 2026-11-14T19:58:41Z, squash e3332dc7; fixes issue #198).** `claimedExit` is no longer defaulted to 0 when failures are reported — a session that honestly reports partial failures (e.g. 34 pass / 7 fail) is no longer misjudged as breach on the exit code alone; when `claimedExit` is null the exit comparison is skipped and only counts are compared.

**Acceptance evidence, all measured and second-party verified by the main round on merged main (e3332dc7), not worker claims.** ① Honest-report replay: 9 recorded replay sessions re-run, 15 checked / 15 confirmed / 0 breach / 0 unknown; the key session live_mwggxca7_e525b2eb flipped from breach to confirmed after the #198 fix, replay exit 1, counts 34/7/41. ② Injected fake agent: all three cases breach — case A (false claim "All 41 tests pass") judged breach (exit: claimed 0, replayed 1) and `--enforce` killed the session on the spot (exit 143); case B (inflated 41/0) and case C (false 33/7, `claimedExit` null taking the counts path) breach, counts path not weakened. ③ Real failing session replay: `npm test` re-run in place, exit 1, 41 tests / 34 pass / 7 fail, matching the agent's honest report and the replay counts exactly. ④ No regression: PR #199 full suite 806 tests, 801 pass, the 4 failures all pre-existing failures reproduced on clean main; merged-main CI run 43634979460 success.

Honest note, stated plainly: ① acceptance criterion ③ changed by user ruling (2026-11-15 01:25): the original "accumulate real breach in the wild" bar is replaced by mechanism evidence — injected fake agent + real failure replay + re-run after the #198 fix. ② Wild real breach is still 0 cases, moved to non-blocking follow-up by the same ruling. claude stream coverage is still in follow-up under issue #187.

Acceptance ledger, stated plainly: the issue #5 ledger is settled as a conclusion on the MiniMax full-matrix recheck (135/135 complete, 2026-11-10) — P2-1 passed, P1-2 / P1-3 / P2-3 did not pass — so no bench numbers are claimed in this release.

## 0.5.34 - Two merged PRs, no milestone acceptance

Patch release. Two merged PRs since 0.5.33, verified end to end in CI and locally — M14 is not accepted yet, so this release carries no milestone claim.
Expand Down
2 changes: 1 addition & 1 deletion docs/launch-dossier.md
Original file line number Diff line number Diff line change
Expand Up @@ -117,7 +117,7 @@ Evidence:
- Adoption records: 1; status: pass
- Goal audit: pass; goal: Keep evalid credible as an open-source agent delivery runtime: repo-owned workflow contracts, executable verification gates, durable project context, and honest benchmark evidence.
- Package evidence: missing; source: registry_json
- Package name/version: @surlymochan/evalid@0.5.34
- Package name/version: @surlymochan/evalid@0.5.35

## Passing Evidence

Expand Down
4 changes: 2 additions & 2 deletions package-lock.json

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

2 changes: 1 addition & 1 deletion package.json
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
{
"name": "@surlymochan/evalid",
"version": "0.5.34",
"version": "0.5.35",
"description": "Claimed != verified: a runtime verification layer for coding agents. It seals each test run and replays it from a clean snapshot, with a zero-judge gate pipeline (spec clarification, code review, TDD proof) and the xflow CLI.",
"keywords": [
"ai-agents",
Expand Down