chore: align governance with shipped artifact types (#62)

Closes #61\n\nImplemented and independently reviewed with AI assistance from Jasper on behalf of Magnus Hedemark.

.github/PULL_REQUEST_TEMPLATE.md

@@ -19,4 +19,6 @@
 
 ## Review notes
 
+- [ ] Final-head review evidence is tied to commit SHA `<sha>`, and all actionable findings are resolved.
+
 <!-- Call out compatibility concerns, intentional simplifications, or follow-up work. -->

.github/workflows/validate.yml

@@ -20,3 +20,11 @@ jobs:
           ruby-version: '3.3'
       - name: Validate skill format and links
         run: ruby scripts/validate-skills.rb
+      - name: Check tracked repository artifacts
+        run: python3 scripts/check-artifacts.py
+      - name: Build tracked Python packages
+        run: |
+          wheel_dir=$(mktemp -d)
+          while IFS= read -r pyproject; do
+            python3 -m pip wheel --no-deps "./$(dirname "$pyproject")" --wheel-dir "$wheel_dir"
+          done < <(git ls-files -- ':(glob)**/pyproject.toml')

CONTRIBUTING.md

@@ -32,6 +32,10 @@ ruby scripts/validate-skills.rb
 
 The same validation runs in GitHub Actions for pushes and pull requests. If a skill includes executable scripts or a package, run its documented checks as well and include the commands and results in your pull request.
 
+## Deprecating a skill
+
+When replacing a skill, preserve its old directory as a routing stub. Prefix its description with `Deprecated: use <replacement>`, explain the migration in the stub, and remove it only after compatibility is no longer required.
+
 ## Pull requests
 
 - Create a branch from `main`: `feat/short-description`, `fix/short-description`, or `docs/short-description`.

README.md

@@ -2,7 +2,7 @@
 
 A collection of AI agent skills — reusable workflows, protocols, and knowledge packs for agentic systems. Skills follow the [Agent Skills open format](https://agentskills.io), making them compatible with any agent framework that supports the standard.
 
-Bundles organize related skills under a single umbrella with shared reference material and auto-loading by trigger context; they appear in the same alphabetical catalog as standalone skills.
+Bundles are this repository's convention for organizing related skills under a single umbrella with shared reference material; they appear in the same alphabetical catalog as standalone skills. Compatible harnesses are guaranteed to see the umbrella skill. Nested subskill auto-loading depends on the harness or on the umbrella skill's instructions.
 
 ## Skills
 

SECURITY.md

@@ -17,4 +17,4 @@ Do not include credentials, private URLs, or personal data in the report unless
 
 You should receive an acknowledgement within seven days. We will investigate, keep the reporter informed when practical, and coordinate disclosure after a fix or mitigation is available.
 
-Supported versions are the latest commit on `main` and the most recent tagged release. Older releases may receive security fixes when the impact warrants it.
+Supported version: the current `main` branch only.

agent-council/README.md

@@ -38,6 +38,8 @@ Or install from the skill directory:
 pip install -e /path/to/agent-council/
 ```
 
+Pip and wheel installs include generated and user-supplied personas, but not the `hermes-profiles` library. To use real bundled profiles, start from a recursive source checkout.
+
 ## Usage
 
 ```bash

agent-council/SKILL.md

@@ -82,7 +82,7 @@ Options:
 
 ## Profile Selection
 
-By default, the council auto-selects relevant profiles from a library of **39 real professional profiles** (shipped as a git submodule from the [hermes-profiles](https://github.com/magnus919/hermes-profiles) repository). Each profile has a SOUL.md — an identity document with real methodology, values, and operating principles — rather than fabricated personas.
+In a recursive source checkout, the council auto-selects relevant real professional profiles from the included [hermes-profiles](https://github.com/magnus919/hermes-profiles) library. Each profile has a SOUL.md — an identity document with real methodology, values, and operating principles — rather than a fabricated persona. Pip and wheel installs do not bundle that library; use generated or user-supplied personas instead.
 
 ### Auto-selection
 
@@ -102,10 +102,10 @@ Three ways to populate the council, with different tradeoffs:
 
 | Method | Best for | Diversity | Setup |
 |--------|----------|-----------|-------|
-| `--profiles` (auto-select) | Single-domain questions with clear keywords | High — profiles have real SOUL.md methodology | Zero — just ask the question |
-| `--profiles name1,name2` | Targeted debates where you know the stakeholders | Highest — you pick specific methodological voices | Know the profile names |
+| `--profiles` (auto-select) | Single-domain questions with clear keywords | High — profiles have real SOUL.md methodology | Recursive source checkout required |
+| `--profiles name1,name2` | Targeted debates where you know the stakeholders | Highest — you pick specific methodological voices | Recursive source checkout and profile names |
 | `--persona-file file.json` | Full control over agent identities, custom domains | Variable — depends on how you design them | Create a JSON file |
-| Auto (no flag) | Default — uses profiles if available, falls back to generated | Good — auto-selects from 39 profiles | Zero |
+| Auto (no flag) | Default — uses profiles if available, falls back to generated | Good — varies with available profiles | No setup for generated personas; recursive source checkout for real profiles |
 
 For most cases, let it auto-select or use `--profiles` with 3-5 names. Only use `--persona-file` when you need specific invented expertise that doesn't map to any existing profile.
 

agent-council/agent_council/phases/select.py

@@ -2,7 +2,6 @@
 
 import json
 import os
-import subprocess
 import sys
 from pathlib import Path
 
@@ -11,24 +10,10 @@ import yaml
 from agent_council.state import ProfileInfo
 
 
-# Path to the profiles submodule within the skill directory
+# Path to the locally available profiles library within the skill directory.
 PROFILES_DIR = Path(__file__).resolve().parent.parent.parent / "profiles" / "profiles"
 
 
-def _update_submodule() -> None:
-    """Pull the latest profiles from the hermes-profiles submodule."""
-    skill_root = PROFILES_DIR.parent  # agent-council/profiles/
-    try:
-        subprocess.run(
-            ["git", "submodule", "update", "--remote", "--init"],
-            cwd=skill_root.parent,  # agent-council/
-            capture_output=True,
-            timeout=30,
-        )
-    except Exception:
-        pass  # Non-fatal — use whatever version we have
-
-
 def _list_available() -> list[str]:
     """List all available profile names."""
     if not PROFILES_DIR.exists():
@@ -63,8 +48,7 @@ def _load_profile(name: str) -> ProfileInfo | None:
 
 
 def load_all() -> list[ProfileInfo]:
-    """Load all available profiles. Updates submodule first."""
-    _update_submodule()
+    """Load all locally available profiles."""
     profiles = []
     for name in _list_available():
         p = _load_profile(name)
@@ -75,7 +59,6 @@ def load_all() -> list[ProfileInfo]:
 
 def select_by_names(names: list[str]) -> list[ProfileInfo]:
     """Load specific profiles by name."""
-    _update_submodule()
     profiles = []
     for name in names:
         name = name.strip().lower()

agent-council/tests/test_select.py

@@ -0,0 +1,43 @@
+import importlib.util
+import sys
+import tempfile
+import types
+import unittest
+from dataclasses import dataclass
+from pathlib import Path
+from unittest.mock import patch
+
+
+yaml = types.ModuleType("yaml")
+yaml.safe_load = lambda _: {}
+sys.modules["yaml"] = yaml
+
+state = types.ModuleType("agent_council.state")
+
+
+@dataclass
+class ProfileInfo:
+    name: str
+    description: str
+    soul_content: str
+
+
+state.ProfileInfo = ProfileInfo
+sys.modules["agent_council.state"] = state
+
+select_spec = importlib.util.spec_from_file_location(
+    "select", Path(__file__).parents[1] / "agent_council" / "phases" / "select.py"
+)
+select = importlib.util.module_from_spec(select_spec)
+select_spec.loader.exec_module(select)
+
+
+class ProfileSelectionTests(unittest.TestCase):
+    def test_missing_profiles_are_empty_without_running_git(self):
+        with tempfile.TemporaryDirectory() as directory:
+            with patch.object(select, "PROFILES_DIR", Path(directory) / "profiles"):
+                self.assertEqual(select.load_all(), [])
+
+
+if __name__ == "__main__":
+    unittest.main()

agent-skills/SKILL.md

@@ -66,7 +66,6 @@ The `SKILL.md` file must contain YAML frontmatter followed by Markdown body cont
 | `compatibility` | No | Max 500 chars. Indicates environment requirements. |
 | `metadata` | No | Arbitrary key-value mapping. |
 | `allowed-tools` | No | Space-separated string of pre-approved tools. (Experimental) |
-| `confirmation` | No | Boolean (default: false). Signals destructive operations requiring explicit user confirmation. |
 
 #### `name` field rules
 - 1–64 characters

agent-skills/references/specification.md

@@ -33,7 +33,7 @@ The `SKILL.md` file must contain YAML frontmatter followed by Markdown content.
 | `compatibility` | No       | Max 500 characters. Indicates environment requirements (intended product, system packages, network access, etc.). |
 | `metadata`      | No       | Arbitrary key-value mapping for additional metadata.                                                              |
 | `allowed-tools` | No       | Space-separated string of pre-approved tools the skill may use. (Experimental)                                    |
-| `confirmation` | No | Boolean (default: false). Signals destructive or state-changing operations requiring explicit user confirmation. |
+
 
 <Card>
   **Minimal example:**
@@ -197,10 +197,6 @@ The optional `allowed-tools` field:
   ```
 </Card>
 
-#### `confirmation` field
-
-The optional `confirmation` field must be a boolean. When `true`, the agent must obtain explicit user confirmation before carrying out destructive or state-changing operations directed by the skill. This is a safety signal, not a permissions system. Most skills should omit it.
-
 ### Body content
 
 The Markdown body after the frontmatter contains the skill instructions. There are no format restrictions. Write whatever helps agents perform the task effectively.

scripts/check-artifacts.py

@@ -0,0 +1,106 @@
+#!/usr/bin/env python3
+"""Validate runnable repository artifacts selected from the Git index."""
+
+import argparse
+import json
+import subprocess
+import sys
+import unittest
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parent.parent
+
+
+def tracked_files() -> list[Path]:
+    result = subprocess.run(
+        ["git", "ls-files", "-z"],
+        cwd=ROOT,
+        check=True,
+        capture_output=True,
+    )
+    return [ROOT / name for name in result.stdout.decode().split("\0") if name]
+
+
+def is_shell_script(path: Path) -> bool:
+    if path.suffix in {".sh", ".bash"}:
+        return True
+    try:
+        first_line = path.open("rb").readline().decode("utf-8", "replace")
+    except OSError:
+        return False
+    return first_line.startswith("#!") and any(
+        shell in first_line for shell in ("/sh", "/bash", "env sh", "env bash")
+    )
+
+
+def test_directories(files: list[Path]) -> list[Path]:
+    directories = set()
+    for path in files:
+        for parent in path.parents:
+            if parent.name == "tests":
+                directories.add(parent)
+                break
+            if parent == ROOT:
+                break
+    return sorted(directories)
+
+
+def run_checks(files: list[Path]) -> list[str]:
+    errors = []
+    for path in files:
+        relative = path.relative_to(ROOT)
+        if path.suffix == ".py":
+            try:
+                compile(path.read_bytes(), str(relative), "exec")
+            except (OSError, SyntaxError) as error:
+                errors.append(f"python {relative}: {error}")
+        elif path.suffix == ".json":
+            try:
+                json.loads(path.read_text(encoding="utf-8"))
+            except (OSError, json.JSONDecodeError) as error:
+                errors.append(f"json {relative}: {error}")
+        if is_shell_script(path):
+            result = subprocess.run(
+                ["bash", "-n", str(path)], cwd=ROOT, text=True, capture_output=True
+            )
+            if result.returncode:
+                errors.append(f"bash -n {relative}: {result.stderr.strip()}")
+
+    for directory in test_directories(files):
+        relative = directory.relative_to(ROOT)
+        result = unittest.TextTestRunner(verbosity=0).run(
+            unittest.defaultTestLoader.discover(
+                str(directory), top_level_dir=str(directory)
+            )
+        )
+        if not result.wasSuccessful():
+            errors.append(
+                f"unittest discover {relative}: "
+                f"{len(result.failures)} failures, {len(result.errors)} errors"
+            )
+    return errors
+
+
+def self_check() -> None:
+    files = [ROOT / "example/tests/test_example.py", ROOT / "elsewhere/file.py"]
+    assert test_directories(files) == [ROOT / "example/tests"]
+    assert is_shell_script(ROOT / "scripts/check-artifacts.py") is False
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--self-check", action="store_true")
+    args = parser.parse_args()
+    if args.self_check:
+        self_check()
+        return 0
+
+    errors = run_checks(tracked_files())
+    for error in errors:
+        print(error, file=sys.stderr)
+    return 1 if errors else 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())