# Awesome Agentic Engineering > Executable AI agent production readiness: a zero-dependency scorecard and > GitHub Actions gate, prompt-injection fixtures, machine-readable result > contracts, risk profiles, incident regressions, and reproducible runtime > adapters. Documentation is maintained in English and Simplified Chinese. Canonical repository: https://github.com/lindixu6-hash/awesome-agentic-engineering ## Start Here - Web scorecard: https://lindixu6-hash.github.io/awesome-agentic-engineering/ - Production readiness guide (English): https://lindixu6-hash.github.io/awesome-agentic-engineering/guide/ - 生产就绪指南（简体中文）: https://lindixu6-hash.github.io/awesome-agentic-engineering/zh/guide/ - Five-minute fail-closed quickstart: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/docs/quickstart.md - 五分钟 fail-closed 快速接入: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/docs/quickstart.zh-CN.md - English README: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/README.md - 中文 README: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/README.zh-CN.md ## Downloadable Fail-Closed Starters Each Agent Card starts at 0/20 with one launch blocker. Each workflow enables strict blocker failure. CI checks byte equality with agentic-init output. - Read-only Agent Card: https://lindixu6-hash.github.io/awesome-agentic-engineering/starters/read-only/agent-card.json - Read-only workflow: https://lindixu6-hash.github.io/awesome-agentic-engineering/starters/read-only/agent-readiness.yml - Draft-only Agent Card: https://lindixu6-hash.github.io/awesome-agentic-engineering/starters/draft-only/agent-card.json - Draft-only workflow: https://lindixu6-hash.github.io/awesome-agentic-engineering/starters/draft-only/agent-readiness.yml - State-changing Agent Card: https://lindixu6-hash.github.io/awesome-agentic-engineering/starters/state-changing/agent-card.json - State-changing workflow: https://lindixu6-hash.github.io/awesome-agentic-engineering/starters/state-changing/agent-readiness.yml ## Machine-Readable Contracts - Schema usage and pinning guide: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/schema/README.md - Schema 使用与版本锁定指南: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/schema/README.zh-CN.md - Agent Card JSON Schema: https://lindixu6-hash.github.io/awesome-agentic-engineering/schema/agent-card.schema.json - Prompt-injection fixture JSON Schema: https://lindixu6-hash.github.io/awesome-agentic-engineering/schema/prompt-injection-fixture.schema.json - Eval Result JSON Schema: https://lindixu6-hash.github.io/awesome-agentic-engineering/schema/eval-result.schema.json - Readiness profile JSON Schema: https://lindixu6-hash.github.io/awesome-agentic-engineering/schema/readiness-profiles.schema.json - Risk-profile catalog: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/profiles/readiness-profiles.json - Eight inert prompt-injection fixtures: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/evals/prompt-injection/fixtures.jsonl ## Executable Adapters - LangGraph prompt-injection eval (English): https://lindixu6-hash.github.io/awesome-agentic-engineering/langgraph-eval/ - LangGraph 提示注入 Eval（简体中文）: https://lindixu6-hash.github.io/awesome-agentic-engineering/zh/langgraph-eval/ - LangGraph.js adapter source: https://github.com/lindixu6-hash/awesome-agentic-engineering/tree/main/adapters/langgraph - OpenAI Agents SDK eval (English): https://lindixu6-hash.github.io/awesome-agentic-engineering/openai-agents-eval/ - OpenAI Agents SDK Eval（简体中文）: https://lindixu6-hash.github.io/awesome-agentic-engineering/zh/openai-agents-eval/ - OpenAI Agents SDK adapter source: https://github.com/lindixu6-hash/awesome-agentic-engineering/tree/main/adapters/openai-agents - Attested evidence provenance: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/docs/evidence-provenance.md - 带 Attestation 的证据溯源: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/docs/evidence-provenance.zh-CN.md - Deterministic reference adapter: https://github.com/lindixu6-hash/awesome-agentic-engineering/tree/main/adapters/reference-runtime ## Evidence - Latest release: https://github.com/lindixu6-hash/awesome-agentic-engineering/releases/latest - Main CI: https://github.com/lindixu6-hash/awesome-agentic-engineering/actions/workflows/ci.yml - Public LangGraph evidence run: https://github.com/lindixu6-hash/awesome-agentic-engineering/actions/runs/31975175069 - Public OpenAI Agents SDK evidence run: https://github.com/lindixu6-hash/awesome-agentic-engineering/actions/runs/31980983499 - Attested producer/verifier run: https://github.com/lindixu6-hash/awesome-agentic-engineering/actions/runs/31981738763 - Attested bundle SHA-256: 81671c0e9589e65413e13b7ca7a19d3453166ae783cb5ae3feb4b46565256521 - Real consumer audit (Content OS): https://github.com/lindixu6-hash/ai-content-workflow-skills/actions/runs/31974318431 - Source-linked production incidents (English): https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/docs/production-incidents.md - 带来源的生产事故（简体中文）: https://github.com/lindixu6-hash/awesome-agentic-engineering/blob/main/docs/production-incidents.zh-CN.md ## Commands Initialize a fail-closed starter: npm exec --yes --package=github:lindixu6-hash/awesome-agentic-engineering#v0 -- agentic-init --profile read-only --name "My Agent" Run the offline OpenAI Agents SDK adapter: npm run install:openai-agents SOURCE_DATE_EPOCH=1786924800 npm run eval:openai-agents Score an existing Agent Card: npm exec --yes --package=github:lindixu6-hash/awesome-agentic-engineering#v0 -- agentic-score agent-card.json Validate generated eval results: npm exec --yes --package=github:lindixu6-hash/awesome-agentic-engineering#v0 -- agentic-validate-results results.jsonl --fixtures fixtures.jsonl ## Interpretation Limits - A score is a reviewed engineering claim, not a benchmark of model capability. - A high total never overrides an explicit launch blocker. - Structural validation proves shape and internal consistency, not artifact authenticity or agent safety. - The bundled runtime policies are deterministic examples, not LLM benchmarks. - LangGraph adapter results do not prove arbitrary LangGraph applications safe. - OpenAI Agents adapter results do not prove arbitrary OpenAI Agents SDK applications safe. - Artifact attestation proves artifact and workflow identity, not semantic correctness, independent review, branch protection, or general Agent safety. - Content OS is a same-maintainer prototype adopter; its passing compatibility score does not hide its failing draft-only profile or unresolved blockers. - Do not infer external production adoption from repository Stars, CI status, examples, or self-operated evidence. ## License MIT. The repository uses synthetic identifiers, reserved `.example` domains, and inert fixtures. It contains no live secret or working exfiltration endpoint.