tegwick a852d3f1ff feat(mvp): working secrets-engine CLI for the whynot-design npm publish lane

Implements SECRETS-WP-0002 end to end as a uv-managed Python package:

- catalog: non-secret lane registry + strict validator (build/test/prod)
- stage roles + OpenBao ACL policies; guards refuse wildcards, sys/, identity/,
  admin names, and cross-stage paths before any backend call
- plan/apply: dry-run-first, idempotent policy + approle apply, decision-gated
- decisions: State Hub lookup with local-fixture fallback; non-secret evidence
  to JSONL + hub progress, scrubbed of any value
- provision/verify: mode-0600 file import + generated test values; positive/
  negative checks that never print the value
- exec delivery: `exec --catalog ... -- npm publish` injects the token via a
  temp .npmrc for the child only, cleaned up on exit/failure/interrupt
- ops-warden routing contract + hardening backlog docs
- 34 tests incl. live OpenBao integration; scripts/demo-e2e.sh runs the full
  chain against a throwaway bao dev server

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2026-06-28 12:28:45 +02:00

2.4 KiB

Raw Blame History

ops-warden → secrets-engine Routing Contract

ops-warden issues SSH certificates only. Every other credential need (API keys, provider tokens, DB passwords, npm publish tokens) routes to secrets-engine, which is OpenBao-backed. ops-warden must never request, hold, cache, or vend a raw secret value. A route result is a pointer, not a key.

What ops-warden calls

secrets-engine route <catalog-id> --json

What it returns (the contract)

{
  "catalog_id":       "whynot-design-npm-publish",
  "owner":            "whynot-design",
  "stage":            "prod",
  "decision_status":  "resolved",
  "decision_ref":     "whynot-design-npm-publish",
  "review_url":       "http://127.0.0.1:8000/decisions/<id>",
  "metadata_applied": true,
  "value_present":    true,
  "ready":            true,
  "next_command":     "secrets-engine exec --catalog whynot-design-npm-publish -- <command...>",
  "missing":          ""
}

Field	Meaning
`decision_status`	`resolved`/`approved` => approved; `missing`/`pending` => not yet
`metadata_applied`	OpenBao ACL policy + approle exist for the lane
`value_present`	the secret value has been provisioned (boolean only — value never read)
`ready`	approved and applied and provisioned
`next_command`	the single safe command the caller should run next
`missing`	the one human/provisioning step still outstanding

Guarantees

No value crosses this boundary. route reports a boolean value_present, derived from a metadata/presence check — it never reads the secret.
Actionable when not ready. If a lane is unapproved, unapplied, or unprovisioned, next_command + missing tell the caller exactly what to do.
Idempotent / read-only. route performs no mutation.

whynot-design retry flow

whynot-design CI needs a publish token → asks ops-warden.
ops-warden runs secrets-engine route whynot-design-npm-publish --json.
If ready=false, it surfaces missing + next_command to the human (e.g. "needs approved decision" or "needs provisioning").
Once ready=true, the workload runs secrets-engine exec --catalog whynot-design-npm-publish -- npm publish.

Anti-patterns (forbidden)

ops-warden POST /messages/ asking for NPM_TOKEN / OPENROUTER_API_KEY.
Caching value_present as if it were the value.
Inventing warden secret / warden bao — they do not exist.

2.4 KiB Raw Blame History