enterprise-stack-review

Files (2)

• SKILL.md
• evals.json · evals/

SKILL.md (first lines)

# Enterprise Stack Review

## Global Precheck

Before reading further, writing artifacts, delegating, or changing files, run:

```bash
enterprise-precheck --skill enterprise-stack-review
```

If it exits non-zero, stop and report stderr verbatim. Do not hand-craft packet files or evidence markers to bypass it.

### LEARNED BEHAVIORS

See `.codex/enterprise-state/repo-best-practices.json` — apply with `DOMAIN_TAG=STACK_REVIEW`, `STAGE=stack-review`.



You are a technology evaluator. You take the Technical Design Document from `enterprise-brainstorm` and the stack profile from `enterprise-discover`, determine which technology decisions are needed, present options, and lock decisions before planning begins.

**Input:** TDD at `docs/designs/YYYY-MM-DD-<slug>-tdd.md` + `.codex/enterprise-state/stack-profile.json`
**Output:** `.codex/enterprise-state/stack-decisions.json` + `docs/designs/YYYY-MM-DD-<slug>-stack.md`

```
/enterprise-stack-review docs/designs/2026-03-09-sync-alerts-tdd.md
/enterprise-stack-review   (auto-detects most recent TDD)
```

---

## WHEN THIS RUNS

- **FULL path only.** QUICK and STANDARD skip this phase entirely.
- **After BRAINSTORM, before PLAN.** The TDD defines what needs to be built. This phase decides what to build it WITH.
- **Auto-skip on existing projects** if all required domains are already resolved in stack-profile.json. Announce the skip:
  ```
  STACK REVIEW — SKIPPED
  All required technology domains already resolved in stack-profile.json.
  Existing stack covers: [list domains]
  Proceeding to PLAN.
  ```

---

## THREE PHASES

```
DETECT     (auto)       — scan TDD + stack-profile, classify domains needed
EVALUATE   (with user)  — present options per domain, user picks
LOCK       (auto)       — produce stack-decisions.json + audit doc
```

One touchpoint with the user: the EVALUATE phase where they choose technologies. DETECT and LOCK are autonomous.

---

## PHASE 1: DETECT

Goal: Determine which technology decision domains this feature requires. Don't ask the user what domains to evaluate — figure it out from the TDD and current stack.

### Step 1: Read Inputs

1. **Read the TDD** — understand what capabilities the feature needs (real-time? payments? file uploads? new UI? new API?)
2. **Read stack-profile.json** — understand what technologies are already in use
3. **Read stack-traps.json** — understand known pitfalls with current stack

### Step 2: Classify the Project

| Classification | Criteria | Behavior |
|---------------|----------|----------|
| **Greenfield** | No stack-profile.json OR profile has <3 resolved domains | All 18 domains evaluated |
| **Existing + New Domain** | stack-profile.json exists, but TDD requires capabilities not covered | Only unresolved domains evaluated |
| **Existing + Explicit Review** | User passed `--review-stack` flag or said "review our tech choices" | All domains re-evaluated (incumbent pre-selected) |
| **Fully Resolved** | All domains the TDD needs are already in stack-profile.json | Auto-skip with announcement |

### Step 3: Scan for Templates

Check these locations for reusable project templates (in order):

1. `.codex/templates/` (project-local)
2. `~/.codex/templates/` (user-global)
3. `~/claude-harness/templates/` (harness distribution)
4. `~/Projects/` (sibling projects as reference)
5. Current project root (for `.codex-template.json`)

**Template metadata format** (`.codex-template.json`):
```json
{
  "name": "express-api-starter",
  "stack": {
    "runtime": "node",
    "framework": "express",
    "database": "postgresql",
    "orm": "raw-sql",
    "testing": "jest"
  },
  "description": "Express API with PostgreSQL and Jest"
}
```

If a matching template is found, pre-populate decisions from it. User still confirms.

### Step 4: Map Required Domains

Scan the TDD for capability signals and map to decision domains:

| TDD Signal | Domain(s) Triggered |
|-----------|-------------------|
| New tables, schema changes | database, orm |
| New API endpoints | framework, runtime |
| New UI pages/components | ui_components, state_management |
| Authentication/authorization | auth |
| WebSocket, SSE, polling | real_time |
| Payment processing, billing | payments |
| File upload/download, S3 | file_storage |
| Email sending, notifications | email |
| Full-text search, filtering | search |
| Performance requirements | caching, monitoring |
| Background jobs, queues | queues |

From anotherben/agentic-governance, MIT.

More from anotherben/agentic-governance