Merge pull request 'docs: add Shell Spawn Hard-Stop Rule for agent workflows (Closes #258)' (#281) from docs/issue-258-shell-spawn-hard-stop into master

docs/llm-workflow-runbooks.md

@@ -285,6 +285,33 @@ explicit control-checkout repair.
 
 Portable wording: [`skills/llm-project-workflow/SKILL.md`](../skills/llm-project-workflow/SKILL.md).
 
+## Shell Spawn Hard-Stop Rule
+
+Symptom: a shell tool call returns `exit_code: -1` with empty stdout/stderr.
+That is an executor spawn failure, not a command failure — the command never
+ran, and retrying the identical call cannot succeed.
+
+Required behavior (fail closed, issue #258):
+
+1. **Probe once.** On the first spawn failure, run one trivial probe
+   (`echo ok` or `pwd`). If the probe also returns `exit_code: -1`, mark
+   shell unavailable for the session.
+2. **Hard-stop at two.** After two consecutive spawn failures, stop all
+   further shell tool use for the session; never retry the same failing
+   spawn. A hundred retries produce a hundred identical failures (session
+   `019f382e`: 100+ tool calls stalled on a trivial encode-and-commit task).
+3. **Emit a recovery report.** The report must direct the operator to:
+   - restart the session,
+   - kill hung background terminals (a hung test runner holding the
+     executor is a known contributor),
+   - prefer MCP-native paths for remaining mutations (for example
+     `gitea_commit_files` under `gitea.repo.commit`) instead of shell.
+4. **No improvised fallbacks.** Shell unavailability never authorizes
+   WebFetch/browser/manual-encoding workarounds (see #260). No shell means
+   stop-and-report.
+
+Doc-contract tests: `tests/test_shell_spawn_hard_stop_docs.py`.
+
 ## Branch worktree isolation
 
 All LLM implementation and review work happens in an isolated branch worktree

gitea_mcp_server.py

@@ -3629,6 +3629,16 @@ _GUIDE_RULES = {
         "small fixes, review fixes, conflicts, emergencies). Main checkout: "
         "read-only inspect, fetch, create worktrees, post-merge stable "
         "update, explicit repair only."),
+    "shell_spawn_hard_stop": (
+        "A shell result of exit_code: -1 with empty stdout/stderr is an "
+        "executor spawn failure, not a command failure. Probe once (echo/pwd); "
+        "if the probe fails, mark shell unavailable for the session. After "
+        "two consecutive spawn failures, hard-stop all shell use — never "
+        "retry the same failing spawn — and emit a recovery report: restart "
+        "the session, kill hung background terminals, prefer MCP-native "
+        "paths (e.g. gitea_commit_files) for remaining mutations. Shell "
+        "unavailability never authorizes WebFetch/browser/manual-encoding "
+        "fallbacks (#258)."),
 }
 
 _COMMON_WORKFLOWS = [

skills/llm-project-workflow/SKILL.md

@@ -115,6 +115,26 @@ The main checkout may only be used for read-only inspection, fetching,
 stable-branch update after merged PRs, creating `branches/` worktrees, or
 explicit control-checkout repair.
 
+## Shell Spawn Hard-Stop Rule
+
+A shell tool result of `exit_code: -1` with empty stdout/stderr means the
+executor failed to spawn — it is not a command failure, and retrying the same
+call cannot succeed.
+
+1. On the first spawn failure, run one trivial probe (`echo ok` or `pwd`).
+   If the probe also fails, mark shell unavailable for the session.
+2. After two consecutive spawn failures, hard-stop all further shell tool
+   use for the session. Never retry the same failing spawn.
+3. Stop and emit a recovery report instead of improvising fallbacks. The
+   report must tell the operator to: restart the session, kill hung
+   background terminals (a hung test runner is a known contributor), and
+   prefer MCP-native paths (for example `gitea_commit_files`) for any
+   remaining mutations.
+
+Retry spirals are a real failure mode (issue #258: 100+ tool calls on a
+trivial encode-and-commit task). The hard-stop is fail-closed: no shell means
+stop-and-report, not workarounds.
+
 ## B. Isolated worktree rule
 
 **Never implement or review in the main checkout** (Global LLM Worktree Rule).

tests/test_operator_guide.py

@@ -133,7 +133,8 @@ class TestControlPlaneGuide(GuideTestBase):
         for key in ("hard_stops", "fail_closed", "head_sha_pinning",
                     "merge_confirmation", "redaction", "separation",
                     "profile_switching", "identity_verification",
-                    "work_selection", "global_worktree"):
+                    "work_selection", "global_worktree",
+                    "shell_spawn_hard_stop"):
             self.assertIn(key, rules)
         self.assertIn("MERGE PR", json.dumps(rules["merge_confirmation"]))
         self.assertTrue(rules["hard_stops"])

tests/test_shell_spawn_hard_stop_docs.py

@@ -0,0 +1,82 @@
+"""Doc-contract checks for the shell-spawn hard-stop rule (#258).
+
+When the shell executor fails to spawn (exit_code: -1 with empty
+stdout/stderr), agents must probe once, mark shell unavailable, hard-stop
+after two consecutive spawn failures, and emit a recovery report instead of
+retry-spiralling. These tests pin that guidance in the runbooks, the portable
+skill doc, and the operator guide so it cannot silently regress.
+"""
+from pathlib import Path
+
+REPO_ROOT = Path(__file__).resolve().parent.parent
+
+RUNBOOK = REPO_ROOT / "docs" / "llm-workflow-runbooks.md"
+SKILL = REPO_ROOT / "skills" / "llm-project-workflow" / "SKILL.md"
+
+
+def _normalize(text: str) -> str:
+    """Collapse whitespace so phrase checks survive markdown line wrapping."""
+    return " ".join(text.split())
+
+
+def _runbook_text() -> str:
+    return _normalize(RUNBOOK.read_text(encoding="utf-8"))
+
+
+def _skill_text() -> str:
+    return _normalize(SKILL.read_text(encoding="utf-8"))
+
+
+def test_runbook_has_shell_spawn_hard_stop_section():
+    text = _runbook_text()
+    assert "## Shell Spawn Hard-Stop Rule" in text
+    for phrase in (
+        "exit_code: -1",
+        "empty stdout/stderr",
+        "trivial probe",
+        "two consecutive spawn failures",
+        "mark shell unavailable",
+        "recovery report",
+    ):
+        assert phrase in text, f"runbook missing phrase: {phrase!r}"
+
+
+def test_runbook_recovery_report_names_required_steps():
+    text = _runbook_text()
+    for phrase in (
+        "restart the session",
+        "kill hung background terminals",
+        "MCP-native",
+    ):
+        assert phrase in text, f"runbook recovery guidance missing: {phrase!r}"
+
+
+def test_runbook_forbids_retry_spirals():
+    text = _runbook_text()
+    assert "never retry the same failing spawn" in text
+
+
+def test_skill_doc_declares_shell_spawn_hard_stop_rule():
+    text = _skill_text()
+    assert "## Shell Spawn Hard-Stop Rule" in text
+    for phrase in (
+        "exit_code: -1",
+        "two consecutive spawn failures",
+        "recovery report",
+    ):
+        assert phrase in text, f"SKILL.md missing phrase: {phrase!r}"
+
+
+def test_operator_guide_rules_include_shell_spawn_hard_stop():
+    import sys
+
+    sys.path.insert(0, str(REPO_ROOT))
+    import gitea_mcp_server
+
+    rule = gitea_mcp_server._GUIDE_RULES["shell_spawn_hard_stop"]
+    for phrase in (
+        "exit_code: -1",
+        "two consecutive",
+        "recovery report",
+    ):
+        assert phrase in rule, f"operator guide rule missing: {phrase!r}"