Composable Agent Skills for shipping software with the discipline of a senior engineering org—shared contracts, typed artifacts, and hard stop conditions.
Works with Cursor, Claude Code, Codex, and other skills.sh-compatible agents.
# List available skills
npx skills add willianbs/skills --list
# Install all skills
npx skills add willianbs/skills
# Install a single skill
npx skills add willianbs/skills --skill engineering-osPoint your agent at the conductor skill and describe the work:
Use engineering-os: add rate limiting to the public API.
Follow the standard path and stop if ADR or spec blocks.
engineering-os classifies intent, picks a path, runs only the skills you need, threads artifacts between them, and stops on Block.
| Path | Flow |
|---|---|
| lite | context → implement → review → PR |
| standard | context → [grill] → spec → [threat] → ADR → [impact/contract] → plan → [migration] → tests → implement → [obs] → review → [security/perf] → quality gate → PR |
| incident | incident command → context → defect analysis → implement → review → quality gate → PR → postmortem |
| architecture | context → mentor/ADR → [grill/threat/impact] → spec → plan |
| review-only | context → [impact/contract] → review → specialists → quality gate? |
Use loop-engineer to babysit CI/PRs or re-run a path until stop criteria.
| Skill | Role |
|---|---|
| engineering-os | Conductor — routes work through the right path |
| model-router | Capability-tier routing (cost vs depth) |
| context-loader | Minimal, budgeted project context |
| guardrails-configurator | Interview + write project engineering rules |
| requirements-grill | Adversarial clarification before specs |
| spec-validator | Spec completeness and testable acceptance criteria |
| adr-enforcer | Architecture Decision Record compliance |
| delivery-planner | Milestones, risks, rollback, and task graph |
| test-strategy-designer | Risk-based test strategy from ACs |
| tdd-executor | Red/green/refactor implementation loops |
| feature-implementer | Approved, minimal-diff implementation |
| defect-analyst | Investigation and root-cause analysis |
| refactoring-advisor | Safe, behavior-preserving refactors |
| code-reviewer | Diff-scoped correctness review |
| quality-gate | Merge/deploy readiness aggregator |
| pr-generator | PR descriptions grounded in the diff |
| engineering-mentor | Tradeoffs and growth guidance |
| Skill | Role |
|---|---|
| threat-modeler | STRIDE/assets threat models before deep audits |
| security-auditor | Threat-aware security assessment |
| performance-auditor | SLO-aware performance assessment |
| change-impact-analyzer | Blast radius across services and clients |
| api-contract-guardian | OpenAPI/proto breaking-change gate |
| migration-surgeon | Safe expand/contract data migrations |
| observability-designer | SLIs, logs, traces, and alert quality |
| ai-eval-harness | Evals for AI features and skills |
| Skill | Role |
|---|---|
| incident-commander | Severity, comms, and war-room loop |
| postmortem-writer | Blameless postmortems with action items |
| loop-engineer | Recurring loops and CI/PR babysitting |
Each skill has its own README.md and SKILL.md.
Skills share one operating contract:
- Typed artifacts —
CONTEXT_PACK→SPEC_STATUS→PLAN→IMPL_REPORT→QG_REPORT→ … - Stop conditions —
Proceed/ProceedWithConditions/Revise/Block - Severity & confidence — portfolio-standard enums, evidence required for High+
- Uniqueness — each skill owns one job; specialists escalate instead of overlapping
Details: docs/SKILL-CONTRACT.md · docs/ARTIFACTS.md
MIT © Will Silveira