refactor: agents and skills move to a client-neutral source that renders per client

`.claude/agents/` was the source of truth, which made every role Claude-Code
shaped. Adding a second client meant rewriting each role in that client's syntax
and maintaining both copies — the drift this scaffold exists to prevent, one layer
up.

Roles and skills now live under `.agents/` and render into each registered
client. `.claude/agents/`, `.claude/skills/` and `.codex/agents/` are generated;
`scripts/sync-agent-integrations.py --check` fails on drift and belongs in CI.

The role metadata is portable rather than vendor-named: `reasoning_tier`
(deep/balanced/fast/vision), `capabilities`, `mutation`, `invocation`, and an
optional `preload_skills`. A client manifest maps those to native syntax and must
declare what it cannot express — `codex.yaml` declares `tier_policy: unsupported`
and its adapters say so in the file, rather than the tier silently evaporating and
leaving the repository to believe it was enforced.

The port is behaviour-preserving where it should be and a fix where it should not.
Every instruction body is byte-identical — the whole diff to `.claude/agents/` is
18 added lines and zero deletions. What changed is frontmatter that was missing:

- four agents (`code-reviewer`, `tdd-guardian`, `dependency-audit`, `pr-creator`)
  declared no `tools:` and therefore inherited the ENTIRE tool pool, so three
  review-only agents could edit and write the code they were reviewing. All eight
  now declare capabilities explicitly.
- the six read-only roles gain a non-editing permission mode, so the constraint is
  enforced by the client rather than by the prompt asking nicely.
- `mutation` is now explicit, which records the two roles that genuinely need to
  write: `pr-creator` (external-write — it pushes a branch and opens a PR) and
  `dependency-audit` (workspace-write — package managers rewrite lockfiles).

`pr-creator` keeps `shell` because opening a PR needs it, but it is now the only
agent here with a write mutation and a declared reason for it, instead of one of
four with unlimited access by omission.

.claude/README.md

@@ -27,7 +27,7 @@ A comprehensive guidance system for Claude Code with strict TDD enforcement, mul
 ~/.claude/
 ├── CLAUDE.md                    # Core principles (~150 lines)
 ├── settings.json                # Model config (opusplan), Claude hooks
-├── agents/
+├── agents/                      # GENERATED from ../.agents/roles — do not edit
 │   ├── tdd-guardian.md          # TDD enforcement (sonnet)
 │   ├── code-reviewer.md         # PR review (opus)
 │   ├── security-scanner.md      # Security checks (sonnet, read-only)
@@ -36,7 +36,7 @@ A comprehensive guidance system for Claude Code with strict TDD enforcement, mul
 │   ├── pr-creator.md            # Automated PR creation (haiku)
 │   ├── release-notes.md         # CHANGELOG generation (haiku, read-only)
 │   └── plan-reviewer.md         # Plan validation (sonnet, read-only)
-├── skills/
+├── skills/                      # GENERATED from ../.agents/skills — do not edit
 │   ├── languages/
 │   │   ├── python/SKILL.md      # pytest, Ruff, FastAPI
 │   │   ├── typescript/SKILL.md  # Vitest, ESLint, React
@@ -191,6 +191,21 @@ echo '~/.claude/hooks/pre-commit-tdd.sh' > .husky/pre-commit
 @plan-reviewer Review docs/plans/active/3.5_business-capabilities.md
 ```
 
+## Editing agents and skills
+
+`.claude/agents/` and `.claude/skills/` are **generated**. The source of truth is
+[`.agents/`](../.agents/README.md), which is client-neutral so the same roles and
+skills also render for Codex (and any client added later) rather than being
+rewritten per tool.
+
+```bash
+python3 scripts/sync-agent-integrations.py          # regenerate
+python3 scripts/sync-agent-integrations.py --check  # fail on drift (use in CI)
+```
+
+Edit `.agents/roles/<name>.md` or `.agents/skills/**`, then run the sync. A direct
+edit to a generated file is reverted by the next run.
+
 ## Customization
 
 ### Project-Specific CLAUDE.md