# Credential and access routing

**Audience:** Codex, Claude Code, Grok, and custodian agents that call **llm-connect**
for inference. Run this check **before** requesting secrets, API keys, SSH access,
login tokens, or database passwords — in any repo, not only `ops-warden`.

ops-warden **issues SSH certificates** (`warden sign`, `cert_command`) **and is the
operator access front door** for every other credential need. For `exec_capable` lanes
(OpenBao reads, key-cape login) `warden access <need> --fetch/--exec` **proxies the fetch
as you** — it runs the owner's tool with your identity and streams the value to you;
ops-warden holds, caches, and logs nothing. For non-exec lanes it points you at the owner.

**Do not** `POST /messages/` to `ops-warden` expecting a secret *value* — a State Hub
reply is always a pointer. The **value comes from the CLI front door** (`warden access`),
run with **your** identity, never from the inbox.

### Lookup (do this first)

```bash
warden route find "<describe your need>" --json     # who owns it (pointer)
warden access "<describe your need>" --json          # how to get it (handoff)
```

`warden access` is the operator front door (WARDEN-WP-0014): it renders the owner,
auth method, path template, command skeleton, and policy-gate status for any need.
For `exec_capable` lanes it can **proxy the fetch as you** (`--fetch`/`--exec`) — it
runs the owner's tool with **your** identity and streams the value to you; ops-warden
never holds, caches, or logs the value. See `wiki/OperatorAccessAssist.md`.

Requires the `warden` CLI from `~/ops-warden` (`uv tool install .` or `uv run warden`).

| Agent runtime | How to orient |
| --- | --- |
| **Codex / Grok** (shell, HTTP State Hub) | `warden route` commands above; inbox `to_agent=ops-warden` is for coordination, not secret vending |
| **Claude Code** (MCP when available) | `get_domain_summary("custodian")` for workstreams; **still** use `warden route` for credential ownership |
| **llm-connect** (inference service) | Never put secret retrieval in prompts; route custody to OpenBao/operator paths surfaced by `warden route` |

### Quick routing table

| I need… | Owner | ops-warden role |
| --- | --- | --- |
| SSH cert (`adm`/`agt`/`atm`) | ops-warden | **Issue** — `warden sign` |
| API key, DB password, provider token | OpenBao (`railiance-platform`) | **Assist** — `warden access <need> --fetch/--exec` proxies as you; OpenBao keeps custody |
| Login / OIDC / MFA | key-cape / Keycloak | **Assist** — `warden access <need> --fetch` runs the login as you |
| Authorization decision | flex-auth | Route only |
| activity-core → issue-core emission | activity-core + issue-core | Route — `warden route show activity-core-issue-sink` |
| SSH tunnel | ops-bridge (+ `cert_command` from warden) | Route only |

### Anti-patterns (do not do these)

- `POST /messages/` to `ops-warden` asking for `ISSUE_CORE_API_KEY`, `OPENROUTER_API_KEY`, etc.
- Inventing `warden secret`, `warden login`, `warden bao`, `warden tunnel` — they do not exist
- Pasting secrets into Git, State Hub, workplans, logs, or chat
- Treating `warden access --fetch` as a *secret store*. It is a transparent conduit
  using **your** identity — it holds nothing. ops-warden as a **standing broker**
  (its own secret-read token, a cache of fetched values) is forbidden; runtime secret
  custody stays in OpenBao, authorization in flex-auth.

### Other capabilities (reuse-surface)

Non-credential capabilities are usually discovered through **reuse-surface** federation
(`reuse-surface` registry / `capability.*` indexes). Credential routing is inlined in
every repo's agent instructions because it is high-frequency, high-risk, and easy to
get wrong.

**Canon:** `~/ops-warden/wiki/CredentialRouting.md` · catalog `~/ops-warden/registry/routing/catalog.yaml`