Preserve live qualification evidence and close finite C08 stop/repair testing

.github/workflows/plananvil-codex-qualification.yml

@@ -11,6 +11,7 @@ on:
         options:
           - smoke
           - c13
+          - c08
           - diagnostics
           - precision
           - recovery
@@ -248,7 +249,7 @@ jobs:
     if: >-
       github.event_name == 'workflow_dispatch' &&
       github.ref == 'refs/heads/main' &&
-      (inputs.mode == 'full' || inputs.mode == 'c13' || inputs.mode == 'recovery')
+      (inputs.mode == 'full' || inputs.mode == 'c13' || inputs.mode == 'recovery' || inputs.mode == 'c08')
     environment: plananvil-codex
     runs-on:
       - self-hosted
@@ -326,7 +327,10 @@ jobs:
             qualification_args+=(--only C13)
           fi
           set +e
-          if [ "${{ inputs.mode }}" = "recovery" ]; then
+          if [ "${{ inputs.mode }}" = "c08" ]; then
+            qualification_args+=(--only C08)
+          fi
+          if [ "${{ inputs.mode }}" = "recovery" ] || [ "${{ inputs.mode }}" = "c08" ]; then
             python3 tools/live_codex_qualification_recovery.py \
               --root "${QUALIFICATION_REPO}" \
               --source-commit "${GITHUB_SHA}" \

.github/workflows/plananvil-tests.yml

@@ -54,6 +54,8 @@ jobs:
         run: git diff --check
 
   distribution:
+    needs: codex-conformance
+    if: always()
     name: distribution and release candidate
     runs-on: ubuntu-latest
     timeout-minutes: 15
@@ -70,6 +72,11 @@ jobs:
           git config --global user.email "plananvil-ci@example.invalid"
           git config --global commit.gpgsign false
           git config --global protocol.file.allow always
+      - name: Require actual CLI conformance in the protected distribution check
+        if: always()
+        env:
+          CONFORMANCE_RESULT: ${{ needs.codex-conformance.result }}
+        run: test "$CONFORMANCE_RESULT" = "success"
       - name: Run distribution and release tests
         run: python -m unittest discover -s tests -v
       - name: Validate prepared C01-C16 evidence templates
@@ -117,7 +124,9 @@ jobs:
           echo "$RUNNER_TEMP/codex-conformance-bin" >> "$GITHUB_PATH"
           echo "PLANANVIL_TEST_CODEX_BIN=$RUNNER_TEMP/codex-conformance-bin/codex" >> "$GITHUB_ENV"
       - name: Verify real CLI lifecycle against deterministic loopback responses
-        run: python -m unittest discover -s tests -p test_qualification_c09_cli.py -v
+        run: |
+          python -m unittest discover -s tests -p test_qualification_c09_cli.py -v
+          python -m unittest discover -s tests -p test_qualification_c08_cli.py -v
       - name: Restore hosted VM namespace policy
         if: always()
         run: |

CHANGELOG.md

@@ -19,8 +19,8 @@ All notable changes to PlanAnvil are documented here.
 - extend deterministic live qualification for C02, C09, C11, C13, C14, and C16 with explicit-only skill policy, current AGENTS precedence evidence, project-scoped SubagentStart semantics, bytecode-free PlanAnvil bootstrap, and real Git signing/hook failure diagnostics;
 - make C12 a deterministic runtime byte-budget probe with redundant `project_doc_max_bytes` enforcement, secret head/tail markers, zero-tool automatic-loading evidence, and outer PlanAnvil full-file hash verification;
 - make C06, C08, and C09 deterministic live probes using the real PlanAnvil PreToolUse/PreCompact/PostCompact hooks, explicit postcondition evidence, low-limit `body_after_prefix` auto-compaction triggers, checkpoint repair, repeated compaction, and post-second-compaction continuation checks;
-- add a diagnostic-only C13 transport mode that keeps `full` strict-ephemeral, retries non-ephemeral only for the recognized parent-thread registration failure, uses an isolated disposable `CODEX_HOME`, requires cleanup/auth-metadata invariants, and verifies real SubagentStart semantics with an outer-generated context proof before any baseline 2.3 decision;
-- promote C13 qualification to baseline 2.3: full qualification remains ephemeral-first but may use a known-error-gated non-ephemeral retry with a synthetic home-scoped `fixture_agent` in disposable `CODEX_HOME`, while the real `SubagentStart` hook remains project-scoped and cleanup/auth invariants remain fail-closed;
+- retain diagnostic-only historical C13 controllers; the active full controller uses the baseline 2.3 ephemeral-first, known-error-gated fallback described below;
+- promote C13 qualification to baseline 2.3: full qualification remains ephemeral-first but may use a known-error-gated non-ephemeral retry with a project-scoped, explicitly declared `fixture_agent` and project-scoped `SubagentStart` hook; disposable `CODEX_HOME` isolates only trust/auth bridging and persistence and cleanup/auth invariants remain fail-closed;
 - require production releases to use a GitHub-verified signed annotated tag whose target is reachable from `main`;
 - fail the production release gate closed when the release worktree is dirty or Git cleanliness cannot be verified;
 - document the controlled self-hosted Codex qualification path and keep the previous sandbox procedure as a manual fallback.
@@ -55,9 +55,18 @@ All notable changes to PlanAnvil are documented here.
 - make the Git-hook probe fixture emit explicit hook diagnostics while preserving fail-closed classification for unrelated commit failures;
 - synchronize golden blind-review fixtures and dependent comparison hashes with the required independent `plan-anvil-reviewer` author role.
 
+### Qualification closure — 2026-09-07
+
+- preserve complete baseline 2.3 live evidence from full run #25, `34060321283`, tested at `d0384f76bc4150d33bb8f51ef5981f3243b3cfb3` with Codex CLI 0.153.4, `gpt-5.6-sol`, Debian 13; all C01–C16 were reproduced;
+- retain the exact source-bound archive, hashes and original limitations; C13 passed via the permitted project-native non-ephemeral fallback;
+- replace the old C08 repaired-path workload with a finite pressure/finish scenario and strict termination checks; the old positive timeout remains in historical evidence and the replacement still requires live confirmation;
+- add C08-only qualification and real-CLI loopback conformance without changing C09/C10/C13 runtime behavior, product payload or live runner security;
+- reset newly materialized template indices to unexecuted package results instead of inheriting historical success labels;
+- validate qualification archive integrity and qualified product identity, and require committed finite C08 completion evidence before production publication.
+
 ### Release status
 
-0.2.0 is code-complete as a release candidate. Production publication remains blocked until protected-`main` administration is enabled and required C01-C16 live Codex evidence is committed as `REPRODUCED`.
+0.2.0 remains a release candidate. Full baseline qualification is recorded; production publication awaits the finite C08 live follow-up, strict release validation and a verified signed annotated tag. No tag or release has been published by the qualification-closure change.
 
 ## [0.1.0] - 2026-07-12
 

README.md

@@ -10,12 +10,13 @@ It **generates and validates a plan but never executes it**. Product implementat
 
 The deterministic generator core, schemas, templates, tests, optional planning agents, defense-in-depth hooks, repository installer/upgrader/uninstaller, release tooling, and a deterministic C01-C16 live-qualification template archive/materializer are implemented.
 
-Release status is **candidate**, not production-ready. Deterministic CI is green, but production publication remains gated on two external steps:
+**Full baseline 2.3 qualification passed:** run [#25](https://github.com/KeyffMS/PlanAnvil/actions/runs/34060321283) reproduced C01–C16 on source commit `d0384f76bc4150d33bb8f51ef5981f3243b3cfb3`, with Codex CLI `0.153.4`, model `gpt-5.6-sol`, and Debian 13. C04 is informational; the other fifteen capabilities gate qualification. The complete sanitized evidence is committed under `capabilities/` and preserved unchanged under `qualifications/34060321283/`.
 
-1. protect `main` with required PR/CI checks (tracked by issue #6);
-2. execute the prepared C01-C16 packages in an authenticated current Codex sandbox and commit required `REPRODUCED` evidence (tracked by issue #7).
+**Release status remains candidate.** The old C08 positive trial proved its narrow unblock assertion but later timed out. Its finite replacement must finish in a new live `mode=c08` run and its evidence must be committed before production publication. C13 was reproduced through the explicit known-error-gated non-ephemeral fallback, not by proving ephemeral spawning works.
 
-The capability contract is defined in `docs/CODEX_CAPABILITY_BASELINE.md`. Deterministic tests and prepared fixtures do not substitute for live Codex evidence.
+See [current qualification and release status](https://github.com/KeyffMS/PlanAnvil/blob/main/docs/QUALIFICATION_STATUS.md) for the precise evidence boundary and next step. Deterministic CI and loopback CLI conformance are regression checks, not substitutes for live model evidence. Production publication also retains protected-main, clean-tree and verified signed annotated-tag requirements.
+
+The capability contract is defined in `docs/CODEX_CAPABILITY_BASELINE.md`.
 
 ## Install into another repository
 
@@ -108,7 +109,7 @@ python tools/release_check.py --candidate
 ## Release and live qualification
 
 - `docs/RELEASE.md` — deterministic gates, tag workflow and publication contract
-- `docs/CODEX_SANDBOX_RUNBOOK.md` — exact remaining C01-C16 sandbox sequence
+- `docs/CODEX_SANDBOX_RUNBOOK.md` — current full and targeted sandbox procedures
 - `capabilities/templates.part*` + `tools/prepare_capabilities.py` — deterministic prepared C01-C16 fixtures/prompts/config/assertions/results/hashes
 
 A production tag is rejected by `.github/workflows/release.yml` until every required capability is `REPRODUCED`.
@@ -121,11 +122,11 @@ A production tag is rejected by `.github/workflows/release.yml` until every requ
 - `docs/RECOVERY_AND_VALIDATION.md` — crash recovery, checkpoint, schema and path-safety guarantees
 - `docs/OPENAI_COMPLIANCE.md` — Codex compatibility decisions
 - `docs/CODEX_CAPABILITY_BASELINE.md` — reproducible capability release gate
-- `docs/CODEX_CAPABILITY_QUALIFICATION_2026-08-28.md` — latest qualification audit
+- `docs/QUALIFICATION_STATUS.md` — current qualification status and historical audit index
 - `docs/INSTALLATION.md` — install/upgrade/uninstall contract
 - `docs/TROUBLESHOOTING.md` — operational recovery guidance
 - `docs/RELEASE.md` — release workflow
-- `docs/CODEX_SANDBOX_RUNBOOK.md` — remaining live qualification procedure
+- `docs/CODEX_SANDBOX_RUNBOOK.md` — live qualification procedure
 
 ## Author
 

capabilities/C01/README.md

@@ -1,11 +1,30 @@
-# C01 — Capability evidence
+# C01 — Repository skill discovery
 
-- Expected behavior: Repository skills are discovered from `.agents/skills`.
 - Source: `DOCUMENTED`
 - Release-gating: `yes`
-- Current result: `BLOCKED`
-- Qualification attempt: `2026-08-28`
-- Documentation check: `PASS` against current official Codex skills documentation.
-- Live blocker: no authenticated Codex runtime or `codex` executable is available in the qualification environment.
+- Current result: `REPRODUCED`
+- Qualification package state: `READY_FOR_LIVE_RUN`
+- Prepared: `2026-08-28`
 
-Do not change the result to `REPRODUCED` until this directory contains the complete sanitized package defined in `capabilities/README.md`.
+## Objective
+
+Verify a repository skill under .agents/skills is discovered from the repository and can be explicitly selected.
+
+## Live metadata to record
+
+Before changing this result to `REPRODUCED`, record the exact Codex version, model slug, OS, permission mode, project trust, fixture commit, setup/cleanup, sanitized observations, evaluation, and hashes. Do not commit transcripts, credentials, private paths, or unrelated repository data.
+
+## Execution
+
+Use `fixture/`, `config/README.md`, `prompt.txt`, and `run-command.txt`. Replace the current BLOCKED `actual.sanitized.json`/`evaluation.json` with the live result, update `capabilities/index.json`, then run `python tools/rehash_capability.py C01` and `python tools/validate_capabilities.py`.
+
+## Live qualification
+
+- Date: `2026-09-06`
+- Codex: `codex-cli 0.153.4`
+- Model: `gpt-5.6-sol`
+- OS: `Debian GNU/Linux 13 (trixie)`
+- Permission mode: `approval=never; sandbox=per-trial; model-tool network disabled`
+- Project trust: `trusted via CLI override for disposable fixture repositories`
+- Source commit: `d0384f76bc4150d33bb8f51ef5981f3243b3cfb3`
+- Result: `REPRODUCED`

capabilities/C01/actual.sanitized.json

@@ -0,0 +1,162 @@
+{
+  "capability_id": "C01",
+  "environment": {
+    "codex_version": "codex-cli 0.153.4",
+    "fixture_commit": "4df35777e3afd6f5b6a9aecfe22793e65bba83ea",
+    "model": "gpt-5.6-sol",
+    "os": "Debian GNU/Linux 13 (trixie)",
+    "permission_mode": "approval=never; sandbox=per-trial; model-tool network disabled",
+    "project_trust": "trusted via CLI override for disposable fixture repositories",
+    "source_commit": "d0384f76bc4150d33bb8f51ef5981f3243b3cfb3"
+  },
+  "observations": [
+    "The explicit activation trial resolved the repository fixture skill and produced the activation proof `FIXTURE_SKILL_ACTIVE`.",
+    "The nested-working-directory trial verified repository identity, discovered the same fixture skill, and produced `FIXTURE_SKILL_ACTIVE`.",
+    "Both trials completed with PASS outcomes, no blocker, and no recorded error events."
+  ],
+  "result": "REPRODUCED",
+  "schema_version": "1.0",
+  "trials": [
+    {
+      "assertions": [
+        {
+          "evidence": "Explicit invocation resolved `fixture-capability` from `.agents/skills/fixture-capability/SKILL.md`; activation proof: FIXTURE_SKILL_ACTIVE.",
+          "name": "Repository skill resolution and explicit activation",
+          "status": "PASS"
+        }
+      ],
+      "blocker": null,
+      "capability_id": "C01",
+      "event_summary": {
+        "completed_command_items": 1,
+        "completed_file_change_items": 0,
+        "error_events": 0,
+        "event_types": {
+          "item.completed": 5,
+          "item.started": 1,
+          "thread.started": 1,
+          "turn.completed": 1,
+          "turn.started": 1
+        },
+        "item_types": {
+          "agent_message": 2,
+          "command_execution": 2,
+          "error": 2
+        }
+      },
+      "git_after": {
+        "branch": "master",
+        "cached_paths": [],
+        "commit_count": 2,
+        "head": "4df35777e3afd6f5b6a9aecfe22793e65bba83ea",
+        "index_tree": "715ef280db9d5ab81405996e12aa2be15877d964",
+        "refs": [
+          "refs/heads/master"
+        ],
+        "status": [],
+        "worktree_count": 1
+      },
+      "git_before": {
+        "branch": "master",
+        "cached_paths": [],
+        "commit_count": 2,
+        "head": "4df35777e3afd6f5b6a9aecfe22793e65bba83ea",
+        "index_tree": "715ef280db9d5ab81405996e12aa2be15877d964",
+        "refs": [
+          "refs/heads/master"
+        ],
+        "status": [],
+        "worktree_count": 1
+      },
+      "observations": [
+        "FIXTURE_SKILL_ACTIVE"
+      ],
+      "outcome": "PASS",
+      "sandbox": "read-only",
+      "trial": "explicit_repository_skill_activation",
+      "trial_name": "explicit_repository_skill_activation"
+    },
+    {
+      "assertions": [
+        {
+          "evidence": "Explicit invocation produced activation proof FIXTURE_SKILL_ACTIVE.",
+          "name": "The fixture skill appears in repository skill discovery.",
+          "status": "PASS"
+        },
+        {
+          "evidence": "Repository identity was verified from the nested working directory, and the repository skill was then discovered and activated.",
+          "name": "Discovery works when Codex starts from a nested working directory inside the same repository.",
+          "status": "PASS"
+        }
+      ],
+      "blocker": null,
+      "capability_id": "C01",
+      "event_summary": {
+        "completed_command_items": 3,
+        "completed_file_change_items": 0,
+        "error_events": 0,
+        "event_types": {
+          "item.completed": 8,
+          "item.started": 3,
+          "thread.started": 1,
+          "turn.completed": 1,
+          "turn.started": 1
+        },
+        "item_types": {
+          "agent_message": 3,
+          "command_execution": 6,
+          "error": 2
+        }
+      },
+      "git_after": {
+        "branch": "master",
+        "cached_paths": [],
+        "commit_count": 2,
+        "head": "4df35777e3afd6f5b6a9aecfe22793e65bba83ea",
+        "index_tree": "715ef280db9d5ab81405996e12aa2be15877d964",
+        "refs": [
+          "refs/heads/master"
+        ],
+        "status": [],
+        "worktree_count": 1
+      },
+      "git_before": {
+        "branch": "master",
+        "cached_paths": [],
+        "commit_count": 2,
+        "head": "4df35777e3afd6f5b6a9aecfe22793e65bba83ea",
+        "index_tree": "715ef280db9d5ab81405996e12aa2be15877d964",
+        "refs": [
+          "refs/heads/master"
+        ],
+        "status": [],
+        "worktree_count": 1
+      },
+      "observations": [
+        "FIXTURE_SKILL_ACTIVE"
+      ],
+      "outcome": "PASS",
+      "sandbox": "read-only",
+      "trial": "nested_working_directory_discovery",
+      "trial_name": "nested_working_directory_discovery"
+    },
+    {
+      "evaluator_event_summary": {
+        "completed_command_items": 1,
+        "completed_file_change_items": 0,
+        "error_events": 0,
+        "event_types": {
+          "item.completed": 3,
+          "item.started": 1,
+          "thread.started": 1,
+          "turn.completed": 1,
+          "turn.started": 1
+        },
+        "item_types": {
+          "agent_message": 2,
+          "command_execution": 2
+        }
+      }
+    }
+  ]
+}

capabilities/C01/config/README.md

@@ -0,0 +1,11 @@
+# C01 sandbox configuration
+
+Use a disposable project-scoped Codex configuration. Record the exact effective configuration in the sanitized result. Do not copy user/global secrets or unrelated settings.
+
+PlanAnvil baseline settings when applicable:
+
+```toml
+[agents]
+enabled = true
+max_concurrent_threads_per_session = 4
+```

capabilities/C01/evaluation.json

@@ -0,0 +1,7 @@
+{
+  "capability_id": "C01",
+  "expected_met": true,
+  "result": "REPRODUCED",
+  "schema_version": "1.0",
+  "summary": "Sanitized trial evidence demonstrates repository skill discovery both explicitly and from a nested working directory."
+}

capabilities/C01/expected.json

@@ -0,0 +1,9 @@
+{
+  "assertions": [
+    "The fixture skill appears in repository skill discovery.",
+    "Discovery works when Codex starts from a nested working directory inside the same repository."
+  ],
+  "capability_id": "C01",
+  "schema_version": "1.0",
+  "title": "Repository skill discovery"
+}

capabilities/C01/fixture/.agents/skills/fixture-capability/SKILL.md

@@ -0,0 +1,6 @@
+---
+name: fixture-capability
+description: Return exactly FIXTURE_SKILL_ACTIVE when explicitly invoked.
+---
+
+Return exactly `FIXTURE_SKILL_ACTIVE`.

capabilities/C01/fixture/README.md

@@ -0,0 +1,5 @@
+# C01 fixture
+
+Create a disposable Git repository from the contents of this directory. Install the PlanAnvil release into it when the capability exercises PlanAnvil itself. Commit the fixture before starting Codex so before/after Git identity can be compared.
+
+Objective: Verify a repository skill under .agents/skills is discovered from the repository and can be explicitly selected.

capabilities/C01/hashes.json

@@ -0,0 +1,15 @@
+{
+  "algorithm": "sha256",
+  "files": {
+    "README.md": "422e8b7912cf13e186c621d5eeae747575e5389e5b135b157596ee4201819565",
+    "actual.sanitized.json": "9042d7e35a10682bff746c4313193baf4d1947dfd814853ca03f78555825aba9",
+    "config/README.md": "2b6beca33571ad108adcdbac9becc82a7fe658cf672305f01f58f4ea54565b4e",
+    "evaluation.json": "12aeecb91deeaef38f633d67f60451a07ba176539a5430cfc16c82c8bafa7ee5",
+    "expected.json": "c41329fbf6892a65b85cfe1166e86d2a3b59b684a9a81601974fca0c54a0cee7",
+    "fixture/.agents/skills/fixture-capability/SKILL.md": "0c4c7985ed4a06a2fa45d0d6a1887fec0f5a59892ae1d0d88317ac1c82a7a4ff",
+    "fixture/README.md": "a3ac65c07ebf67442b31cf3bc1d28ac5c849bd08fee9cba93b0a31339e7602d8",
+    "prompt.txt": "14542f39efe7d9d60d6993f2da2ce19ff021fb50d499d9a60a2f7d7ea6f2030a",
+    "run-command.txt": "0683eb1cb9ff624f8923fce722ffdd05ea4c592509cf2d25c0879b43e8fb56af"
+  },
+  "schema_version": "1.0"
+}