From 37c3c49297d0dfe46bcc469bf1a212a794c1da01 Mon Sep 17 00:00:00 2001 From: Arye Kogan Date: Mon, 6 Jul 2026 17:31:32 +0300 Subject: [PATCH 1/2] docs: record p05 real forge smoke rerun --- AGENTS.md | 4 +- README.md | 9 +- docs/README.md | 4 +- .../target-state-implementation/README.md | 14 +-- .../2026-07-06-p05-real-forge-smoke-rerun.md | 97 +++++++++++++++++++ docs/design/evidence/README.md | 1 + 6 files changed, 117 insertions(+), 12 deletions(-) create mode 100644 docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md diff --git a/AGENTS.md b/AGENTS.md index 5fe3d5c..7a9add2 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -46,7 +46,9 @@ or makes a public SDK/provider/MCP stability promise. The real execution-host path is now selectable on macOS and proves an honest `weak` `process-group` posture at compose time; the current real GitHub Forge/GitHub Issues path includes blocked-PR surfacing, held-merge replay safety, landing-path redaction, and origin-bearing work-source provenance — -held-merge and commit-status legs are unit-proven, with real-effect evidence pending a rerun. +the success-path real Forge `open-pr` smoke has been rerun on the current checkout, while +held-merge and commit-status/comment block-surfacing legs remain unit-proven with real-effect +evidence still owed. Strong no-phone-home evidence remains future work. EVRUN-partial is recorded; P11 also records a blocked EVRUN-full capture attempt where the narrow Codex app-server and real-host smokes passed outside the Codex sandbox, but the combined real GitHub path could not run without sandbox diff --git a/README.md b/README.md index 79b55d4..885ae75 100644 --- a/README.md +++ b/README.md @@ -40,10 +40,11 @@ Current evidence proves a scoped real-provider path with a scripted agent leg real-host smokes pass outside the Codex sandbox, but the combined real GitHub path could not run without sandbox credentials and an integrity key. The current real GitHub Forge/GitHub Issues path also carries blocked-PR surfacing, held-merge replay safety, landing-path redaction, and -per-candidate origin-bearing provenance — held-merge and commit-status legs are unit-proven, with -real-effect evidence pending a rerun. `EVRUN-full`, remote execution, public package publication, -and a full Codex-driven agent leg remain future work; strong/no-phone-home execution-host evidence -remains in that later proof step. +per-candidate origin-bearing provenance — the success-path real Forge `open-pr` smoke has been +rerun on the current checkout, while held-merge and commit-status/comment block-surfacing legs +remain unit-proven with real-effect evidence still owed. `EVRUN-full`, remote execution, public +package publication, and a full Codex-driven agent leg remain future work; strong/no-phone-home +execution-host evidence remains in that later proof step. ## Quick Start diff --git a/docs/README.md b/docs/README.md index ebdfdee..ad706c2 100644 --- a/docs/README.md +++ b/docs/README.md @@ -25,7 +25,9 @@ scoped to **Jig the execution engine**. Current implementation truth includes a private setup command that emits validated owner configuration, plus a private real GitHub Forge/GitHub Issues path with blocked-PR surfacing, held-merge replay safety, landing-path redaction, and origin-bearing candidate provenance — -held-merge and commit-status legs are unit-proven, with real-effect evidence pending a rerun; +the success-path real Forge `open-pr` smoke has been rerun on the current checkout, while +held-merge and commit-status/comment block-surfacing legs remain unit-proven with real-effect +evidence still owed; phase-table PR-number status updates still live in delivery docs. Preserve active design history in place: diff --git a/docs/delivery/target-state-implementation/README.md b/docs/delivery/target-state-implementation/README.md index 454e0ed..6080e81 100644 --- a/docs/delivery/target-state-implementation/README.md +++ b/docs/delivery/target-state-implementation/README.md @@ -345,12 +345,14 @@ boundary enforcement from P02 on, docs checks, and the definition of "delivered" Per the escalation rule (product owns what/why; design reconciles to product; delivery names conflicts rather than resolving them): -- **`MERGE-5` is partially proven; real-effect rerun is still owed.** The product guarantees - blocked-PR surfacing. Surfacing is now reachable from normal runner wiring, and the PR-creation - and comment legs were smoke-proven (`#64`). The commit-status write/read leg failed with an - HTTP 403, and held-merge posture was never observed against a protected target — the - EVRUN-partial record explicitly does not claim `MERGE-5` end to end. A real-effect rerun remains - required before the claim is fully proven; see the review record +- **`MERGE-5` is partially proven; block-surfacing real-effect evidence is still owed.** The + product guarantees blocked-PR surfacing. Surfacing is now reachable from normal runner wiring, + and the success-path real Forge `open-pr` smoke was rerun on the current checkout + (`docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md`). The commit-status/comment + block-surfacing leg remains unit-proven, the earlier commit-status write/read attempt failed + with an HTTP 403, and held-merge posture was never observed against a protected target — the + EVRUN-partial-style smoke records explicitly do not claim `MERGE-5` end to end. Additional + real-effect proof remains required before the claim is fully proven; see the review record `docs/archive/reviews/2026-07-06-p04-p12-target-state-review.md` for the observed failure detail. - **GUARD-2 pause shape is an open design question** (distinct sub-state vs. reusing `parked`, diff --git a/docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md b/docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md new file mode 100644 index 0000000..46c2f06 --- /dev/null +++ b/docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md @@ -0,0 +1,97 @@ +--- +title: "2026-07-06 - P05 real Forge smoke rerun" +status: applied - coordinator privileged run 2026-07-06 +date: 2026-07-06 +--- + +# 2026-07-06 - P05 Real Forge Smoke Rerun + +## Probe + +This reruns the EVRUN-partial real-provider smoke against the disposable sandbox repository +`agentic-workflow-kit/jig-smoke-target` from the current post-PR #77 Jig checkout. The smoke +exercises the real GitHub Issues work-source, real GitHub Forge `open-pr` landing path, Jig run +records, and HMAC integrity sidecar with a scripted agent leg. + +The coordinator supplied the GitHub token and a fresh integrity key only in the command +environment. The emitted evidence facts are the bounded, non-secret JSON written by +`packages/jig-sdk/tests/smoke/evrun-partial.smoke.test.ts` via `EVRUN_EVIDENCE_OUT`; local +`run.json` and `integrity.json` captures stayed under ignored temp directories and are cited by +content hash. + +## Tool Version Pins + +| Tool | Version | +| ----------------- | ---------------------------------------------------------------------- | +| Node.js | `v26.4.0` | +| pnpm | `11.9.0` | +| GitHub CLI (`gh`) | `2.87.2 (2026-02-20)` | +| Git | `2.53.0` | +| Jig checkout | `a78f67caad3c5826cfd02e59a16447c72467256f` (`main` after PR #77 merge) | + +## Inputs + +- Smoke repository: `agentic-workflow-kit/jig-smoke-target` +- Seed issue: `https://github.com/agentic-workflow-kit/jig-smoke-target/issues/1` +- Seed issue label: `jig-candidate` +- Changed file in the sandbox target: `evrun-partial-smoke.txt` +- Landing action: `open-pr` +- Opt-in command shape: + +```bash +EVRUN_SMOKE=1 \ +JIG_GITHUB_ISSUES_REPOSITORY=agentic-workflow-kit/jig-smoke-target \ +JIG_GITHUB_ISSUES_LABEL=jig-candidate \ +EVRUN_EVIDENCE_OUT=/private/tmp/jig-p05-real-forge-smoke-evidence.json \ +corepack pnpm exec vitest run packages/jig-sdk/tests/smoke/evrun-partial.smoke.test.ts +``` + +Credential-bearing variables (`GITHUB_TOKEN`, `GH_TOKEN`, `JIG_RECORDS_INTEGRITY_KEY`) were +present only in the coordinator command environment and are intentionally omitted. + +## Run Summary + +| Field | Value | +| ------------------------- | ------------------------------------------------------------------------------------------ | +| Run ID | `run-evrun-partial-smoke-1783347978055-d5849612-3836-4fff-82ec-b6ef2cc77d18` | +| Pull request URL | `https://github.com/agentic-workflow-kit/jig-smoke-target/pull/12` (closed during cleanup) | +| Landing families observed | `runner-action.opened-pr` | +| Final status | `success` | +| Integrity verification | `verified` | +| Test result | `1` smoke test passed | + +## Content Hashes + +| Artifact | SHA-256 | +| ---------------- | ------------------------------------------------------------------ | +| `run.json` | `4534c0e328bd875001c1584b318a9d7f3f4964d4fc8b0fd5f585b8c3c0aefa47` | +| `integrity.json` | `857069eb8ec7f7cd63cd252d7b45fd585f7d10bf72b8735ada3ec493d6ff1740` | + +## Cleanup + +The smoke-created sandbox pull request was closed after capture and its branch was deleted: +`https://github.com/agentic-workflow-kit/jig-smoke-target/pull/12`. + +## Redaction + +No tokens or credentials appear in this record. The sandbox repository name is intentionally +cited. The smoke asserts the serialized run artifacts and emitted evidence facts do not include +the GitHub token, `x-access-token`, the askpass path, credential-bearing URLs, or the integrity +key. The integrity sidecar carries only the digest and the non-secret key id supplied for the run. + +## Supported IDs + +- Refreshes P05 real Forge/work-source success-path evidence on the current post-PR #77 + checkout. +- Supports that a real GitHub Issues candidate can reach scheduling, a real GitHub Forge + `open-pr` landing effect can occur, and Jig records plus verifies the integrity sidecar. +- Supports the status-doc distinction that this path is real-effect evidence, not merely unit + coverage. + +## Limitations + +This remains an EVRUN-partial-style smoke with a scripted agent leg. It does not prove +EVRUN-full, real Codex editing, real execution-host confinement, adversarial no-phone-home +behavior, branch-protection or merge-queue hold posture, commit-status/comment block surfacing, +or repeated landing idempotency against the live adapter. `MERGE-5` and held-merge real-effect +proof therefore remain open. diff --git a/docs/design/evidence/README.md b/docs/design/evidence/README.md index cbe5dac..acb814e 100644 --- a/docs/design/evidence/README.md +++ b/docs/design/evidence/README.md @@ -27,6 +27,7 @@ guarantee, invariant, or decision IDs the evidence supports. | 2026-07-04 | [N1A-P11/P12 resume and durability](./2026-07-04-n1a-p11-p12-resume-durability.md) | Persistent thread resume and durable app-server session-state evidence for N1b. | | 2026-07-04 | [EVRUN partial real-provider smoke evidence](./2026-07-04-evrun-partial-smoke.md) | Applied evidence for one `work-source -> forge -> records-integrity` smoke run against the disposable sandbox, with a scripted/injected agent leg and explicit EVRUN-full limitations. | | 2026-07-06 | [EVRUN-full capture attempt](./2026-07-06-evrun-full-capture-attempt.md) | P11 capture attempt showing successful narrow Codex app-server and real-host smokes, plus the exact missing-prerequisite blocker that keeps EVRUN-full unproven. | +| 2026-07-06 | [P05 real Forge smoke rerun](./2026-07-06-p05-real-forge-smoke-rerun.md) | Current-checkout rerun of the real GitHub Issues -> GitHub Forge `open-pr` smoke against the disposable sandbox, with records integrity verified and explicit MERGE-5 limitations. | ## N1a deferred probes From 79ae9d57bfdd72811fa6d30ca8baba8d5d7e363d Mon Sep 17 00:00:00 2001 From: Arye Kogan Date: Mon, 6 Jul 2026 17:37:06 +0300 Subject: [PATCH 2/2] docs: narrow p05 smoke redaction claim --- .../design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md b/docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md index 46c2f06..6687e87 100644 --- a/docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md +++ b/docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md @@ -76,8 +76,9 @@ The smoke-created sandbox pull request was closed after capture and its branch w No tokens or credentials appear in this record. The sandbox repository name is intentionally cited. The smoke asserts the serialized run artifacts and emitted evidence facts do not include -the GitHub token, `x-access-token`, the askpass path, credential-bearing URLs, or the integrity -key. The integrity sidecar carries only the digest and the non-secret key id supplied for the run. +the GitHub token, `x-access-token`, the askpass path, or credential-bearing URLs. The integrity +key was supplied only as a coordinator environment variable; the integrity sidecar carries the +HMAC digest and the non-secret key id supplied for the run, not the raw key. ## Supported IDs