菜单
Run the paiOS quality gate suite: cargo fmt, clippy, tests, Taskmaster task validation, and a quick architecture sanity check. Reports what passes, what fails, and what to fix. Use for "check code health", "run quality gates", or after a batch of implementation tasks.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Execute a paiOS Taskmaster m0 task end-to-end: load context from the linked GitHub issue and docs, implement strictly within the declared write-set, run the verify command, and mark the task done. Use whenever the user says "arbeite an Task N" or "implement task N".
Drive the autonomous paiOS M0 loop: repeatedly select the next eligible loop-friendly Taskmaster task, delegate context-load / implement / review to subagents at complexity-matched models, run the quality gate, commit per task, and mark it done. Skips the decision-heavy backlog. Use for "start the loop", "run the autonomous loop", "arbeite die m0-Tasks autonom ab".
Review a paiOS code change (diff, PR, or named files) against the project's architecture standards: hexagonal ports, feature-flag discipline, Rust style, ADR compliance, and test coverage. Returns a prioritised finding list. Use for "review this change", "check this PR", or "review task N output".
| name | paios-code-health |
| description | Run the paiOS quality gate suite: cargo fmt, clippy, tests, Taskmaster task validation, and a quick architecture sanity check. Reports what passes, what fails, and what to fix. Use for "check code health", "run quality gates", or after a batch of implementation tasks. |
Run all quality gates and report results. All commands run from engine/ unless noted.
cd engine && cargo fmt --all -- --check
If it fails: run cargo fmt --all to fix, then re-check. Report which files were reformatted.
cd engine && cargo clippy --all-targets -- -D warnings
Report all warnings. Do not suppress with #[allow(...)] unless the reason is documented inline.
cd engine && cargo test
Report failing tests with their name and the first 10 lines of output. Do not mark passing if any test fails.
Check that key profiles compile without error:
cd engine && cargo build --features desktop 2>&1 | tail -3
For rockchip profile (if aarch64 target is installed):
cd engine && cargo build --features rockchip --target aarch64-unknown-linux-gnu 2>&1 | tail -3
python3 .taskmaster/scripts/validate_tasks.py .taskmaster/tasks/tasks.json
Report any tasks missing required sections (Load context / Acceptance / Verify).
Check for cross-crate import violations:
grep -r "use inference::\|use api::" engine/crates/core/src/ 2>/dev/null && echo "VIOLATION: core imports domain crate" || echo "OK: core has no domain imports"
grep -r "use api::" engine/crates/inference/src/ 2>/dev/null && echo "VIOLATION: inference imports api" || echo "OK"
grep -r "use inference::" engine/crates/api/src/ 2>/dev/null && echo "VIOLATION: api imports inference directly" || echo "OK (should go via core ports)"
Gate 1 fmt: PASS / FAIL (N files reformatted)
Gate 2 clippy: PASS / FAIL (N warnings)
Gate 3 tests: PASS / FAIL (N/M tests, N failing)
Gate 4 desktop: PASS / FAIL
Gate 5 tasks: PASS / FAIL (N issues)
Gate 6 arch: PASS / FAIL (N violations)
Overall: PASS (all green) | FAIL (fix gates above before next task)