# Shared output schema Operational skills that produce user-facing results must follow this baseline. Skill-specific sections extend these rules. They do not replace them. **User-facing results:** Write outcome text, structured sections, and next steps in pragmatic Simplified Technical English (STE). Prefer short sentences, active voice, and approved modals (`can`, `must`, `will`). Do not use `should`, `would`, `may`, `might`, or `could`. Do not use contractions. Prefer `make sure that` over `ensure`, `verify`, or `confirm` as verbs. ## Triggers (frontmatter) - **Strong:** concrete situation + artifact ("fix CI failure", "wire domain X"). - **Weak:** generic "help with code" — route via project entry doc or skills index. - **Not this skill:** list explicit handoffs to adjacent skills when boundaries overlap. ## Preflight Before you run a skill that needs external tools, make sure that the tool is present, or state the fallback: | Tool | Typical use | If missing | | ----------------- | ----------------------------------- | ---------------------------------------------------------------------- | | Docker | backend dev/test, integration tests | Scope to file lint/typecheck. State that backend validation is blocked | | `gh` | PR/issue workflows | File output only. Skip `gh pr edit` when `gh` is missing | | Issue tracker MCP | issue `start`/`finish` | Fail on `start`. Warn on `finish` | ## Final response shape Every operational skill ends with **at least one** of: 1. **Outcome block** — what was done or decided (2–5 sentences max). 2. **Structured section** — skill-specific headings (see each skill). 3. **Next steps** — bullets with skill handoffs when work continues elsewhere. Use repo-relative paths in backticks. Prefer imperative leads for action items. ## Validation defaults For code changes, prefer scoped validation on touched paths. Avoid full-repo lint/check for small edits.