@misterhuydo/sentinel 1.5.31 → 1.5.32

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@misterhuydo/sentinel",
-  "version": "1.5.31",
+  "version": "1.5.32",
   "description": "Sentinel — Autonomous DevOps Agent installer and manager",
   "bin": {
     "sentinel": "./bin/sentinel.js"

package/python/sentinel/__init__.py

@@ -1 +1 @@
-__version__ = "1.5.31"
+__version__ = "1.5.32"

package/python/sentinel/repo_task_engine.py

@@ -34,7 +34,7 @@ from .git_manager import _git_env
 
 logger = logging.getLogger(__name__)
 
-_META_PREFIXES = ("REPO:", "TYPE:", "SUBMITTED_BY:", "SUBMITTED_AT:", "RUN_AT:", "NOTIFY:", "ORIGIN_CHANNEL:")
+_META_PREFIXES = ("REPO:", "TYPE:", "SUBMITTED_BY:", "SUBMITTED_AT:", "RUN_AT:", "NOTIFY:", "ORIGIN_CHANNEL:", "SKIP_TESTS:")
 _TASK_TIMEOUT = 900  # 15 minutes
 
 
@@ -51,6 +51,7 @@ class RepoTask:
     timestamp: str = ""
     run_at: datetime | None = None  # UTC; task is held until this time if set
     origin_channel: str = ""       # Slack channel ID where task was requested (DM or group)
+    skip_tests: bool = False        # skip test suite (safe for trivial/non-logic changes)
 
     def __post_init__(self):
         if not self.fingerprint:
@@ -62,6 +63,20 @@ class RepoTask:
 
 def _build_repo_prompt(task: RepoTask, repo: RepoConfig) -> str:
     submitted = f"Requested by: <@{task.submitter_user_id}>" if task.submitter_user_id else ""
+    if task.skip_tests:
+        test_instruction = (
+            "4. The user explicitly requested to skip tests — syntax check only, "
+            "do NOT run the test suite."
+        )
+    else:
+        test_instruction = (
+            "4. After making your changes, run: git diff HEAD  to review the diff, then decide:\n"
+            "   - If ALL changes are non-logic (log statements, comments, string literals,\n"
+            "     config values, whitespace/formatting): syntax check only — skip the test suite.\n"
+            "   - If ANY logic changed (conditionals, data flow, new/removed code paths,\n"
+            "     method behaviour): run the full test suite (mvn test / npm test / gradlew test)\n"
+            "     and commit only if it passes."
+        )
     return (
         f"You are implementing a requested change in the repository at {repo.local_path}.\n"
         f"Repository: {repo.repo_name}\n"
@@ -76,7 +91,7 @@ def _build_repo_prompt(task: RepoTask, repo: RepoConfig) -> str:
         f"   before making any changes.\n"
         f"2. Implement the requested change following the repo's existing code style.\n"
         f"3. Syntax/compile check modified files.\n"
-        f"4. Run tests if available (Maven, Gradle, npm test) — commit only if passing.\n"
+        f"{test_instruction}\n"
         f"5. Commit all changes:\n"
         f"   git add -A\n"
         f"   git commit -m \"{task.task_type}(<scope>): <concise summary> [sentinel-task]\"\n"
@@ -347,6 +362,7 @@ def drop_repo_task(
     notify_user_ids: list | None = None,
     run_at: datetime | None = None,
     origin_channel: str = "",
+    skip_tests: bool = False,
 ) -> Path:
     """Drop a repo task file into <project_dir>/repo-tasks/.
 
@@ -372,6 +388,8 @@ def drop_repo_task(
         lines.append(f"NOTIFY: {','.join(notify_user_ids)}")
     if origin_channel:
         lines.append(f"ORIGIN_CHANNEL: {origin_channel}")
+    if skip_tests:
+        lines.append("SKIP_TESTS: true")
     lines += ["", description]
     fpath.write_text("\n".join(lines), encoding="utf-8")
     if run_at:
@@ -404,6 +422,7 @@ def scan_repo_tasks(project_dir: Path) -> list[RepoTask]:
         notify_user_ids: list = []
         run_at: datetime | None = None
         origin_channel = ""
+        skip_tests = False
         body_start = 0
 
         for i, line in enumerate(lines):
@@ -434,6 +453,9 @@ def scan_repo_tasks(project_dir: Path) -> list[RepoTask]:
             elif upper.startswith("ORIGIN_CHANNEL:"):
                 origin_channel = stripped[15:].strip()
                 body_start = i + 1
+            elif upper.startswith("SKIP_TESTS:"):
+                skip_tests = stripped[11:].strip().lower() in ("true", "yes", "1")
+                body_start = i + 1
             elif any(upper.startswith(p) for p in _META_PREFIXES) or not stripped:
                 body_start = i + 1
             else:
@@ -469,6 +491,7 @@ def scan_repo_tasks(project_dir: Path) -> list[RepoTask]:
             notify_user_ids=notify_user_ids,
             run_at=run_at,
             origin_channel=origin_channel,
+            skip_tests=skip_tests,
         ))
         logger.info("Found repo task: %s → %s (type=%s)", f.name, repo_name, task_type)
 

package/python/sentinel/sentinel_boss.py

@@ -116,6 +116,9 @@ BEFORE CALLING dev_task OR repo_task — GATHER A COMPLETE SPEC:
 Ask follow-up questions until you have enough to write an unambiguous task description.
 Typical questions to resolve:
 - For repo_task: which repo exactly? (confirm if multiple match)
+- For repo_task: Claude Code auto-judges whether to run tests by inspecting its own diff.
+  Only set skip_tests=true when the user explicitly says "skip tests", "don't run tests",
+  or similar — never set it on your own judgment.
 - What exactly should happen? (specific behaviour, not vague intent)
 - Any config values, credentials, or external dependencies involved?
 - How should it be triggered? (schedule, event, API call, etc.)
@@ -532,6 +535,10 @@ When to act vs. when to ask:
 - Explaining a tool ("what does X do?") → explain naturally, then offer to run it if relevant.
 - NEVER gate investigation on user approval. If diagnosing a problem, run all relevant read tools
   first, then present findings. Asking "Want me to look?" wastes a round trip.
+- NEVER answer from session memory alone when the question is about current state (commits,
+  tasks, fixes, releases). Always call get_status or list_recent_commits first to verify live
+  state. Session memory is a snapshot — tasks complete, commits land, queues drain between turns.
+  If you remember "task X was in-flight", check whether it finished before telling the user to wait.
 - Prefer filter_logs over search_logs when synced logs are available — it's instant and never causes session timeout.
   Use search_logs only when the user explicitly wants live/real-time data or synced logs are not yet available.
 - If a tool call will take a moment (search, fetch, pull), prefix your reply with a brief "working" line ending in "..." before the results, e.g. "Searching SSOLWA for TryDig activity..." then the actual output.
@@ -916,6 +923,15 @@ _TOOLS = [
                         "If omitted the task runs immediately on the next poll cycle."
                     ),
                 },
+                "skip_tests": {
+                    "type": "boolean",
+                    "description": (
+                        "Only set to true when the user explicitly requests skipping tests "
+                        "(e.g. 'skip tests', 'don\\'t run tests', 'just commit it'). "
+                        "When false (default), Claude Code automatically decides whether to run "
+                        "tests by inspecting its own diff — non-logic changes skip tests on their own."
+                    ),
+                },
             },
             "required": ["repo_name", "description"],
         },
@@ -2473,6 +2489,8 @@ async def _run_tool(name: str, inputs: dict, cfg_loader, store, slack_client=Non
         if not _project_dirs:
             return json.dumps({"error": "No project directory found."})
 
+        skip_tests = bool(inputs.get("skip_tests", False))
+
         from .repo_task_engine import drop_repo_task as _drop_repo_task
         task_file = _drop_repo_task(
             _project_dirs[0],
@@ -2483,6 +2501,7 @@ async def _run_tool(name: str, inputs: dict, cfg_loader, store, slack_client=Non
             notify_user_ids=notify_ids,
             run_at=run_at_dt,
             origin_channel=channel,
+            skip_tests=skip_tests,
         )
         logger.info(
             "Boss repo_task: dropped %s for user %s (repo=%s, run_at=%s)",