fix(ci): read the test shard jobs in the Bun crash census and fix stale CI test job docs

.ai/skills/hack-repo-verify/SKILL.md

@@ -98,7 +98,7 @@ untracked candidate additions. Keep private artifacts excluded from the index.
 | Native authored command slices | `tests/native-authored-command.test.ts` (unsupported matrix), `tests/native-authored-command-dispatch.test.ts`, `tests/native-authored-project-{start,run,observe,logs,exec,recovery}.test.ts` and the [command contract](../../../docs/reference/native-authored-commands.md); the dispatch "source CLI" cases need the prepared compiler (`bun run build:config-compiler` or absolute `HACK_CONFIG_COMPILER_BINARY`) and fail without it; `native-authored-project-start` and `native-authored-project-exec` are nondeterministic on a local macOS host, so a single local pass or failure is not proof; live acceptance runs host-serialized on one prepared macOS provider slot |
 | Native VM/routing/reclamation | [candidate guide](../../../docs/guides/native-candidate.md), relevant acceptance gate and isolated host fixture; Linux checks cannot qualify macOS effects |
 | Resource/performance claims | [performance skill](../hack-repo-performance/SKILL.md), matched workload and measured boundaries |
-| Bun runtime crashes | Read-only history: `bun run ci:bun-crash-census -- --since <UTC> --until <UTC>` (failed `test` job logs; `--scan-passing` for all; since the suite moved to `test shard i/4` jobs, pass `--job "test shard i/4"` once per shard for newer runs). Matched reproduction: the gated [experiment](../../../.github/workflows/bun-crash-experiment.yml), which spends hosted macOS minutes and needs an explicit go-ahead; never loop on a shared host. Its default `focused` mode loops `tests/native-https-owner.test.ts` (where both original segfaults happened); `full` needs a budget of at least 40 minutes because one full suite takes 30-35. A run never starts unless its estimate fits the remaining budget, so read `summary.json` for the verdict and per-mode crash, failure and skipped-for-budget counts. Both share [the classifier](../../../scripts/lib/bun-crash-classify.ts) |
+| Bun runtime crashes | Read-only history: `bun run ci:bun-crash-census -- --since <UTC> --until <UTC>` (failed test-suite job logs: every `test shard i/n` job, or the `test` job of a run from before the split, since `test` is now the logless fan-in; `--scan-passing` for all; `--job NAME` selects by exact name or a `*` pattern such as `"test shard *"`; a `warning:` line means no job matched or a selected job ran no Bun, so its zero is not evidence). Matched reproduction: the gated [experiment](../../../.github/workflows/bun-crash-experiment.yml), which spends hosted macOS minutes and needs an explicit go-ahead; never loop on a shared host. Its default `focused` mode loops `tests/native-https-owner.test.ts` (where both original segfaults happened); `full` needs a budget of at least 40 minutes because one full suite takes 30-35. A run never starts unless its estimate fits the remaining budget, so read `summary.json` for the verdict and per-mode crash, failure and skipped-for-budget counts. Both share [the classifier](../../../scripts/lib/bun-crash-classify.ts) |
 
 Read [CI](../../../.github/workflows/ci.yml) for current hosted gates. A local pass
 does not mean CI ran.

docs/reference/shared-host-lifecycle.md

@@ -226,9 +226,10 @@ Progress events are also closed: `lock-waiting`, `lock-acquired`, `claim-stale`,
 
 `bun test tests/shared-host-lifecycle.test.ts` runs portable local fixtures. Each
 fixture is a Bun process on a loopback port with a private HOME. No Docker, real
-HOME, keychain or privileged effect is involved. The file runs in the hosted macOS
-`test` job and is registered in the hosted Linux `linux-process-lifetime` job. A
-Linux pass is a receipt only once that hosted job has run it.
+HOME, keychain or privileged effect is involved. The file runs in one of the hosted
+macOS `test shard i/4` jobs (the required `test` check passes only when every shard
+passes) and is registered in the hosted Linux `linux-process-lifetime` job. A Linux
+pass is a receipt only once that hosted job has run it.
 
 | Scenario | Asserted behavior |
 | --- | --- |

docs/security/env-secrets-threat-model.md

@@ -96,7 +96,7 @@ refer to `hack-dance/hack`.
 | A key copy ships in build output or a static site. | `hack doctor` reports `env-key-public-path` for copies under `public/`, `static/`, `dist/`, `build/`, `out/`, `www/`, `_site/`, `.next/` and `.output/` (#309). |
 | A weak, short or repetitive key. | `env-key-wrong-length` and `env-key-low-entropy` (#309). Generated keys are 32 random bytes in base64url. |
 | Materialized `.hack/.env` or local overrides are committed. | Hack owns a committed `.hack/.gitignore` block, and `hack doctor` flags tracked generated files (see [ignore rules](../env.md#ignore-rules-committed-hackgitignore)). |
-| A secret leaks into a public artifact through a path nobody listed. | `tests/env-secret-sentinel-scan.test.ts` runs in the CI `test` job. It gives a primary checkout and a linked worktree random synthetic secrets through `env add --secret --value-stdin` (shared and `--local`), runs materialize, exec, up (Compose generation), env apply, logs, list, explain and doctor, then commits everything with `git add -A`. It requires zero occurrences of any value or the project key, raw or hex/base64/base64url encoded, in CLI output, recorded child argv, tracked files, every Git object, HACK_HOME, the private HOME and its caches, and generated Compose and runtime files. Plaintext may exist only in `.hack/.env`, and key text only in key files; each must be git-ignored, untracked and mode 0600. `.hack/.internal/compose.env.override.yml` must hold `${HACK_ENV_...}` references and no value, and must also be private and ignored. Values may reach a child's environment only for the exec'd command and `docker compose`. Positive controls prove delivery and that the scanner finds planted encodings in files and Git objects. Docker is a recording shim; no engine, keychain or installed Hack is reached. |
+| A secret leaks into a public artifact through a path nobody listed. | `tests/env-secret-sentinel-scan.test.ts` runs in one of the hosted CI `test shard i/4` jobs; the required `test` check passes only when every shard passes. It gives a primary checkout and a linked worktree random synthetic secrets through `env add --secret --value-stdin` (shared and `--local`), runs materialize, exec, up (Compose generation), env apply, logs, list, explain and doctor, then commits everything with `git add -A`. It requires zero occurrences of any value or the project key, raw or hex/base64/base64url encoded, in CLI output, recorded child argv, tracked files, every Git object, HACK_HOME, the private HOME and its caches, and generated Compose and runtime files. Plaintext may exist only in `.hack/.env`, and key text only in key files; each must be git-ignored, untracked and mode 0600. `.hack/.internal/compose.env.override.yml` must hold `${HACK_ENV_...}` references and no value, and must also be private and ignored. Values may reach a child's environment only for the exec'd command and `docker compose`. Positive controls prove delivery and that the scanner finds planted encodings in files and Git objects. Docker is a recording shim; no engine, keychain or installed Hack is reached. |
 | A dependency vulnerability ships in Hack. | `bun install --frozen-lockfile`, and the read-only `scripts/dependabot-inventory.ts` lists open alerts (#309). Hack does not sandbox project dependencies that run under `hack host exec`; they receive the injected env by design. |
 
 ### Agent and MCP exfiltration