fix: stabilization pass — fail-closed auth, ai flag validation, lint/config/security hardening, regression tests

17-point code-level audit executed via Explorer->Architect->Builder->Tester->Reviewer;
Reviewer accepted (APPROVE_WITH_NOTES; 3 block-list items resolved):

- security: telegram sender-owner AND-gate + TELEGRAM_OWNER_ID, matrix
  MATRIX_ROOM_ID fail-closed, gpg --passphrase-fd 3 (no argv secret),
  /dev/tcp positional-arg form (checkport/smb-client/share-lib/NET_PROBE),
  eval deny-by-default + --no-command-execution carried by both chat bridges,
  tty-gated --trust; config/{telegram,matrix}.env reference templates
- ai: all ExecStart flags validated against installed llama.cpp
  (requested->error, default->omit+warn, CONFIG_REQUESTED_FLAGS); single-file
  hf download failure rc=1 + no .hf-meta; LLAMACPP_HOST coherent;
  POS_SUBCMDS + metadata gaps closed
- tooling: lint-conventions Bash-native rewrite (~24-30x faster, rules and
  output byte-identical, :num restored); pos system uninstall covers all 12
  libs + scale-tail + flags dir + systemd user units (|| true) + plugin
  markers; anchored .bash_completion/.bashrc removal replaces sed -i '/pos/d'
- config: canonical load_env_file in lib/config-ui.sh (CRLF strip, env-wins,
  XDG, LOADED_ENV_KEYS); 9 tools migrated; entertainment-lib collapsed to
  wrappers; docker-compose deliberately unmigrated (source semantics)
- tests: first committed regression suite — tests/run-tests.sh zero-dep
  runner + make test; 12 files / 179 checks / 0 skip / ~52s; hard skip
  contract; systemd-analyze verify on generated unit PASS

Verified: make gen idempotent; make check green; make lint 0 FAIL, 0 WARN;
make test green; bash -n clean; git diff --check clean. Audit deliverables +
agent reports + AGENT_TODO Done entry included.

tests/README.md

@@ -0,0 +1,54 @@
+# tests/ — regression test suite
+
+Zero-dependency Bash regression tests for the Linux_post_install repository.
+
+## Run
+
+```bash
+make test          # discover tests/t-*.sh, run everything
+./tests/run-tests.sh                     # same
+./tests/run-tests.sh t-telegram-auth.sh  # run one file
+```
+
+Exit code is non-zero if any check failed. The suite is designed to be
+deterministic: tests stub every external dependency they touch (curl, gpg,
+sudo, systemctl, llama-server, nvidia-smi…) and run against sandbox temp
+dirs. No network, no sudo, no system changes.
+
+## Adding a test
+
+1. Create `tests/t-<what>.sh` with a `run_test()` function. `set -euo pipefail`
+   is already active (the runner re-asserts it); `$ROOT` is the repo root,
+   `$TEST_TMP` a per-test temp dir that is cleaned automatically.
+2. Use the helpers in `tests/test-lib.sh` for every assertion:
+   `check`, `check_eq`, `check_rc`, `check_contains`, `check_not_contains`,
+   `check_file_exists`, `check_file_absent`, `test_run` / `test_run_env`,
+   `mksandbox`, `tracked_tree_copy`, `count_token`.
+3. If a case cannot run in the current environment (missing binary, missing
+   analyzer), call `skip_case "<desc>" "<reason>"` — never fake a pass and
+   never silently return.
+4. Keep total suite runtime under 90 seconds.
+
+## Skip contract (hard)
+
+A test that cannot run must say `[SKIP] reason`. The runner counts skips in
+the summary and a file whose `run_test()` produced zero checks and zero skips
+is reported as FAIL ("no assertions") — a broken harness can never pass
+silently.
+
+## Suite contents
+
+| File | What it verifies |
+| --- | --- |
+| `t-ai-server-flags.sh` | `pos-ai-server` ExecStart flag set: defaults, CLI, config, dedupe, one-token-per-flag |
+| `t-ai-hf-download.sh` | `pos ai hf download` success + failure honesty (no `.hf-meta` on partial failure, rc != 0) |
+| `t-ai-llama-detect.sh` | `pos ai-server status` version detection, "unknown", missing-binary failure |
+| `t-unsupported-flags.sh` | unsupported-option handling: CLI/config/env hard errors, dropped defaults, word-boundary match |
+| `t-systemd-unit.sh` | generated unit: one ExecStart, quoted paths, `systemd-analyze verify` |
+| `t-telegram-auth.sh` | Telegram listener chat+owner gate and owner-unset fail-closed |
+| `t-matrix-auth.sh` | Matrix listener room+owner gate and room-unset fail-closed |
+| `t-gpg-password.sh` | backup passphrase on fd 3 (never argv), plaintext/corrupt cleanup |
+| `t-config-precedence.sh` | `load_env_file` contract + CLI > env > file > defaults across tools |
+| `t-uninstall-manifest.sh` | install.sh ↔ POS_LIBS symmetry, user-unit discovery, marker-driven plugin removal |
+| `t-gen-docs-drift.sh` | `make gen` idempotence on a pristine tracked tree (CI drift gate) |
+| `t-lint-gate.sh` | `make lint` green on the real tree; planted violations are caught and named |