# /qe:check
Audit the current repository for Agentty quality-enforcement readiness. Do not edit
files, run mutating git commands, create commits, or change repository state. Return a
concise markdown report that lists each rule as `Pass`, `Warn`, or `Unknown`, explains
the evidence you found with repository-root-relative paths, and gives concrete
remediation as a prioritized list of actionable High and Medium severity items.
Apply these rules:
- Existing agent instruction files, such as `AGENTS.md`, `CLAUDE.md`, and `GEMINI.md`,
give agents clear, current, and non-duplicative guidance for the repository. Review
whichever of these files exist without requiring a predefined file set or layout.
- Agent instruction files describe the project purpose, important entry points,
mandatory workflow rules, test expectations, documentation sync expectations, and
validation workflow at the right level of detail.
- Multiple agent instruction files, when present, stay consistent with one another and
avoid near-identical copies that can drift unless the duplication is clearly
intentional and maintained.
- `.pre-commit-config.yaml` exists when the repository has automated quality gates, and
the documented validation workflow points agents to repository-defined hooks instead
of ad hoc commands.
- Agent instructions document per-turn validation requirements, test coverage
expectations, feature-test requirements for user-visible behavior, and the rule to
preserve unrelated user changes when those expectations apply to the repository.
- Architecture documentation exists for repositories with layered or boundary-heavy
code. Check whether the repository documents module ownership, runtime flow,
testability boundaries, and change-path guidance in an appropriate location for that
project.
- Areas that need specialized local guidance have appropriately scoped local agent
instructions, while instruction files avoid exhaustive inventories and focus on
purpose, invariants, entry points, change routing, and docs-sync notes.
Use this report shape:
```markdown
## /qe:check
### Findings
- `Pass` — <rule name>: <repository-root-relative evidence>
- `Warn` — <rule name>: <repository-root-relative evidence>
- `Unknown` — <rule name>: <what could not be verified>
### Action Items
#### High
- <imperative action item that fixes the highest-impact gap>
#### Medium
- <imperative action item for a useful but less urgent improvement>
### Suggested Next Prompt
<one focused prompt the user can send to address the highest-severity action item>
```