Build
When to use this group
Use these skills to prepare, implement, review, and revise work. Commit story documents before using the run engine. Inspect findings and tests before landing.
Choose a skill
| Skill | Result and use |
|---|---|
e11-agent-dev | Prepare implementation, tests, and code review with the developer. |
e11-checkpoint-preview | Review a change with a person and select checks to run. |
e11-code-review | Review code through several methods and triage findings. |
e11-correct-course | Assess a scope change and propose revisions to the initiative. |
e11-customize | Write or audit overrides for installed skills. |
e11-dev-auto | Perform one unattended development iteration. |
e11-dev-flow | Turn a story, fix, or change request into working code. |
e11-epic | Supervise work described by an epic. Use the run CLI for execution. |
e11-errors | Turn production errors into reproduced, staged bugfix stories and publish the fixes, by hand or nightly. |
e11-qa-generate-e2e-tests | Add automated end-to-end tests for existing features. |
e11-retrospective | Review delivered work and record follow-up actions. |
e11-review | Review code or documents for correctness, gaps, structure, and wording. |
e11-spec-flow | Create, update, or validate a specification and its companion documents. |
e11-sprint | Plan the initiative board, report status, or check readiness. |
e11-visual | Show a change as before / after / design-reference images with numbered callouts (optional add-on). |
e11-visual-reconcile | Judge built screens against their design reference and fix, report, or return the drift. |
Use e11-sprint for planning, status, or readiness over the initiative board. Use e11-dev-flow for planning or implementation by conversation. For run supervision, use e11 run and e11 runs.
e11-errors can also run nightly on your own machine: e11 errors nightly pulls, sweeps and runs one capped engine run per staged cause within the night's budget, all in its own worktree beside the checkout, ports each accepted fix to a branch on the PR base, and pushes the reviewed ones with draft PRs itself; agent sessions hold the source token and forge credentials, so they can read what a diagnosis needs from the tracker. To set it up, run e11 errors init <project> --schedule and put the job's environment (source token, forge token, package-registry token, PATH additions) in the nightly.env it names under Foundry Home. See errors.
Review provides seven lenses: adversarial, edge-case-hunter, verification-gap, reachability, alloy-auditor, structure, and prose. Conditions select relevant lenses. Prose follows structure when both apply.