fix: sync lean managed-scope Hermes updates

.github/CODEOWNERS

@@ -0,0 +1,2 @@
+# Default review ownership for Collar
+* @logohere

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,26 @@
+---
+name: Bug report
+about: Report a reproducible bug
+labels: bug
+---
+
+## What happened?
+
+## Expected behavior
+
+## Reproduction steps
+
+1.
+2.
+3.
+
+## Environment
+
+- Collar version:
+- OS:
+- Branch/commit:
+
+## Notes
+
+- Keep reports focused and reproducible.
+- If this touches agent behavior, mention the relevant `.dag` path.

.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,18 @@
+---
+name: Feature request
+about: Propose a small, focused improvement
+labels: enhancement
+---
+
+## What problem are you solving?
+
+## Proposed change
+
+## Why this scope?
+
+## Alternatives considered
+
+## Notes
+
+- Prefer updating existing systems over adding new layers.
+- If this affects agent behavior, include the `.dag` path involved.

.github/pull_request_template.md

@@ -0,0 +1,11 @@
+# Pull request template
+
+PRs must use the repo template:
+
+- `.github/pull_request_template.md`
+
+Keep it lean:
+- explain why the change exists
+- note `.dag` / dotdog impact
+- show tests or verification
+- call out any intentional scope cuts

CONTRIBUTING.md

@@ -4,10 +4,12 @@ DAG-first. `Entity→Target:verb(card)`.
 
 ## Rules
 
-1. **NEVER read `.dog`** — agents query `.dag` via the router
-2. **DAG-path notation** — all behavioral guidance in compact format
-3. **Compile before commit** — run `dag-regen.sh --all` after any `.md`/`.dog` change
-4. **Test harness fallback** — simulate fresh install, verify constants load from `dags/`
+1. **Query `.dag` first** — agent behavior comes from `.dag` routes, not prose
+2. **Keep dotdog in the loop** — compile/spec changes through the dotdog path
+3. **Use the PR template** — every PR should use `.github/pull_request_template.md`
+4. **DAG-path notation** — all behavioral guidance in compact format
+5. **Compile before commit** — run `dag-regen.sh --all` after any `.md`/`.dog` change
+6. **Test harness fallback** — simulate fresh install, verify constants load from `dags/`
 
 ## Project Structure
 

README.md

@@ -88,6 +88,24 @@ collar setup         # first-time setup wizard
 
 - [dotdog](https://github.com/specdog/dotdog) — `.dog` spec format + compiler
 - [specdog](https://github.com/specdog) — the org
+- PRs use an agentic template: `.github/pull_request_template.md`
+
+## PR system
+
+Collar PRs should stay lean and agentic:
+
+- use `.github/pull_request_template.md`
+- explain the reasoning, not just the diff
+- prefer updating existing systems over adding new ones
+- keep `.dag` as the first-read source for agent behavior
+- avoid adding skills or layers that inflate token use unless the change clearly needs them
+- include tests or verification for behavior changes
+
+## Governance
+
+- Branch protection on `main` requires PR review and a passing check
+- Security issues should go through `SECURITY.md`
+- Bugs and features have issue templates in `.github/ISSUE_TEMPLATE/`
 
 ## Dogfood
 

SECURITY.md

@@ -0,0 +1,25 @@
+# Security Policy
+
+## Supported Versions
+
+Security fixes land on the default branch and current release line.
+
+## Reporting a Vulnerability
+
+Do not open a public issue for security bugs.
+
+Use one of these instead:
+- GitHub Security Advisory for this repository
+- Direct contact with the maintainers through the private security channel
+
+Include:
+- affected version or commit
+- impact
+- reproduction steps
+- any suggested mitigation
+
+## Review Expectations
+
+- prefer minimal, targeted fixes
+- avoid adding new systems unless the fix requires them
+- keep `.dag`/dotdog paths lean when changing agent behavior

dag_cli/main.py

@@ -531,7 +531,17 @@ def _resolve_sudo_user_profile_env(name: str) -> str | None:
     if _cfg_path.exists():
         with open(_cfg_path, encoding="utf-8") as _f:
             _early_cfg_raw = _yaml_early.safe_load(_f) or {}
-        if "DAG_REDACT_SECRETS" not in os.environ:
+        # Managed scope: overlay administrator-pinned values so a managed
+        # security.redact_secrets / network.force_ipv4 wins here too. This early
+        # bridge reads config.yaml directly (before load_config is usable), so
+        # without the overlay a managed redact_secrets toggle would be ignored.
+        # Fail-open via the shared helper.
+        try:
+            from hermes_cli import managed_scope
+            _early_cfg_raw = managed_scope.apply_managed_overlay(_early_cfg_raw)
+        except Exception:
+            pass
+        if "DAG_REDACT_SECRETS" not in os.environ and "HERMES_REDACT_SECRETS" not in os.environ:
             _early_sec_cfg = _early_cfg_raw.get("security", {})
             if isinstance(_early_sec_cfg, dict):
                 _early_redact = _early_sec_cfg.get("redact_secrets")

dag_cli/send_cmd.py

@@ -276,6 +276,14 @@ def _load_dag_env() -> None:
     except Exception:
         pass
 
+    # Managed scope: overlay administrator-pinned values before bridging to env,
+    # so a managed top-level scalar wins here too. Fail-open via the helper.
+    try:
+        from hermes_cli import managed_scope
+        raw = managed_scope.apply_managed_overlay(raw if isinstance(raw, dict) else {})
+    except Exception:
+        pass
+
     if not isinstance(raw, dict):
         return
 

gateway/run.py

@@ -1104,6 +1104,15 @@ def _reload_runtime_env_preserving_config_authority() -> None:
             cfg = _yaml.safe_load(f) or {}
         from dag_cli.config import _expand_env_vars
         cfg = _expand_env_vars(cfg)
+        # Managed scope: keep administrator-pinned values authoritative on every
+        # turn too. This per-turn reload re-bridges config→env, so without the
+        # overlay a managed agent.max_turns / timezone / redact_secrets would be
+        # replaced by the user's value after the first turn. Fail-open.
+        try:
+            from hermes_cli import managed_scope
+            cfg = managed_scope.apply_managed_overlay(cfg)
+        except Exception:
+            pass
     except Exception:
         return
 
@@ -1126,6 +1135,17 @@ def _reload_runtime_env_preserving_config_authority() -> None:
         # Expand ${ENV_VAR} references before bridging to env vars.
         from dag_cli.config import _expand_env_vars
         _cfg = _expand_env_vars(_cfg)
+        # Managed scope: overlay administrator-pinned values BEFORE bridging to
+        # env vars, so a managed timezone / redact_secrets / max_turns / terminal
+        # setting wins over the user's value at the env layer too. This bridge
+        # reads config.yaml directly (not via load_config), so without the
+        # overlay every HERMES_*/TERMINAL_* env var below would carry the user's
+        # value even when an administrator pinned it. Fail-open via the helper.
+        try:
+            from hermes_cli import managed_scope
+            _cfg = managed_scope.apply_managed_overlay(_cfg)
+        except Exception:
+            pass
         # Top-level simple values (fallback only — don't override .env)
         for _key, _val in _cfg.items():
             if isinstance(_val, (str, int, float, bool)) and _key not in os.environ:

skills/github/github-pr-workflow/templates/pr-body-bugfix.md

@@ -1,35 +1,42 @@
 ## Bug Description
 
-<!-- What was happening? -->
-
-Fixes #
+- 
 
 ## Root Cause
 
-<!-- What was causing the bug? -->
+- 
 
 ## Fix
 
-<!-- What does this PR change to fix it? -->
+- 
 
--
+## Why this fix
 
-## How to Verify
+- Why this approach?
+- Why this scope?
+- Why not a larger refactor?
+
+## DAG / spec impact
 
-<!-- Steps a reviewer can follow to confirm the fix -->
+- [ ] Uses existing `.dag` / dotdog systems
+- [ ] No new skill unless required
+- [ ] Keeps token overhead low
+- [ ] Updates existing behavior instead of adding layers
+
+## How to Verify
 
 1.
 2.
 3.
 
 ## Test Plan
 
-- [ ] Added regression test for this bug
+- [ ] Added regression test
 - [ ] Existing tests still pass
 - [ ] Manual verification of the fix
 
 ## Risk Assessment
 
-<!-- Could this fix break anything else? What's the blast radius? -->
-
-Low / Medium / High — <!-- explanation -->
+- Impact:
+- Rollback:
+- Notes:

skills/github/github-pr-workflow/templates/pr-body-feature.md

@@ -1,33 +1,40 @@
 ## Summary
 
-<!-- 1-3 bullet points describing what this PR does -->
+- 
 
--
+## Why this change
 
-## Motivation
+- Why now?
+- What user or upstream need does it address?
+- Why this scope and not more?
 
-<!-- Why is this change needed? Link to issue if applicable -->
+## DAG / spec impact
 
-Closes #
+- [ ] Uses existing `.dag` / dotdog systems
+- [ ] No new skill unless required
+- [ ] Keeps token overhead low
+- [ ] Updates existing behavior instead of adding layers
 
-## Changes
+If any box is unchecked, explain why.
 
-<!-- Detailed list of changes made -->
+## Changes
 
--
+- 
 
-## Test Plan
+## Reasoning
 
-<!-- How was this tested? Checklist of verification steps -->
+- What was the design choice?
+- What alternative was rejected?
+- What did you intentionally leave out to keep Collar lean?
 
-- [ ] Unit tests pass (`pytest`)
-- [ ] Manual testing of new functionality
-- [ ] No regressions in existing behavior
-
-## Screenshots / Examples
+## Test Plan
 
-<!-- If UI changes or new output, show before/after -->
+- [ ] Unit tests pass
+- [ ] Manual verification completed
+- [ ] No regression in existing behavior
 
-## Notes for Reviewers
+## Risk
 
-<!-- Anything reviewers should pay special attention to -->
+- Impact:
+- Rollback:
+- Notes: