# Readiness Scouts

Source: https://flow-next.dev/subagents/readiness-scouts/

The nine read-only scouts /flow-next:prime dispatches in parallel to assess whether a repository is ready for agents and production - one scout per pillar.

`/flow-next:prime` answers a single question: **is this repository ready for agents to work in it effectively, and ready for production?** It does that by dispatching nine specialized scouts in parallel: eight each own one *pillar* of readiness, and `docs-gap-scout` checks documentation currency. Every scout is read-only - `Edit`, `Write`, and `Task` are disallowed - so priming a repo never changes it; it only reports.

Prime reports on every pillar but only fixes agent-readiness gaps (with your approval); production-readiness findings are surfaced for you to act on.

## How they fan out

```mermaid
flowchart TB
  Prime["/flow-next:prime"] --> Scouts
  subgraph Scouts["Nine scouts (parallel)"]
    Build["build-scout"]
    Test["testing-scout"]
    Tool["tooling-scout"]
    Md["claude-md-scout"]
    Env["env-scout"]
    Sec["security-scout"]
    Obs["observability-scout"]
    Flow["workflow-scout"]
    Docs["docs-gap-scout"]
  end
  Scouts --> Report["Readiness report"]
```

## Agent-readiness pillars

These directly determine whether an agent can work in the repo without guessing.

| Scout             | Tier           | Pillar                                                                                                     |
| ----------------- | -------------- | ---------------------------------------------------------------------------------------------------------- |
| `build-scout`     | Fast scout     | Build system, scripts, and CI - can an agent compile and run the project to verify its changes?            |
| `testing-scout`   | Fast scout     | Test framework, coverage configuration, and the commands to run tests for fast local verification.         |
| `tooling-scout`   | Fast scout     | Linting, formatting, type-checking, and pre-commit - the fast-feedback loop that catches errors before CI. |
| `claude-md-scout` | Thinking scout | Quality and completeness of `CLAUDE.md` / `AGENTS.md` - conventions, commands, and what *not* to do.       |
| `env-scout`       | Fast scout     | Environment setup: `.env` templates, Docker, devcontainers, and undocumented system dependencies.          |

## Production-readiness pillars

Not strictly required for agents to work, but important context that prime reports.

| Scout                 | Tier       | Pillar                                                                                                      |
| --------------------- | ---------- | ----------------------------------------------------------------------------------------------------------- |
| `security-scout`      | Fast scout | Branch protection and rulesets, CODEOWNERS, and dependency-update configuration (checked via the `gh` API). |
| `observability-scout` | Fast scout | Structured logging, distributed tracing, metrics, and health endpoints.                                     |
| `workflow-scout`      | Fast scout | CI/CD pipelines, PR and issue templates, and workflow automation.                                           |

## Shared with planning

| Scout            | Tier           | Job                                                                                                                                           |
| ---------------- | -------------- | --------------------------------------------------------------------------------------------------------------------------------------------- |
| `docs-gap-scout` | Thinking scout | Flags documentation that has drifted from the code. Also dispatched by [`/flow-next:plan`](https://flow-next.dev/subagents/planning-scouts/). |

The seven mechanical, evidence-gathering scanners are **fast scout** dispatches: a `fast scout` line in your routing block moves them to a cheaper model, and without one they run on the agent’s default (Reasoning tier). `claude-md-scout` and `docs-gap-scout` are **thinking scout** dispatches, because documentation-quality judgment degrades on a fast model. Tiers map to a concrete model per platform - see [the model-tier note on the overview](https://flow-next.dev/subagents/overview/). The judgment of what to fix lives in the `prime` skill itself.