autonomousguy
Open-source AI agent skills for embedded automotive engineers - AUTOSAR, MISRA C, ISO 26262 - that give Copilot and Claude the domain context they are missing.
Install
# Install all skills (global)
npx skills add ptsilivis/autonomousguy -g
# Browse what's inside before installing
npx skills add ptsilivis/autonomousguy --list
That is all you need. Works with Claude Code, GitHub Copilot, Cursor, Codex, Gemini CLI, and any tool that supports Agent Skills.
What’s inside
10 mode-aware skills covering the ECU software lifecycle. Each skill routes from your input to the right sub-workflow, so one skill handles several related tasks. Start with codebase-analysis on a new project: it writes .autonomousguy/CODEBASE_MAP.md, which the other skills reference.
The AUTOSAR column states which platform a skill assumes. Skills marked “Classic + Adaptive” default to Classic AUTOSAR (CP) and switch to Adaptive AUTOSAR (AP) when your input names AP concepts (ara::com, ara::exec, C++14, POSIX, service-oriented, manifest).
| Skill |
What it does |
Platform |
codebase-analysis |
Scans the repo and maps SWCs, BSW usage, port interfaces, RTE API calls, signal flow, and ASIL zones into CODEBASE_MAP.md. |
Classic + Adaptive |
autosar-swc |
Component design, port interface definition, SWC code, plain-text diagrams, and integration review. |
Classic + Adaptive |
autosar-bsw |
BSW config (Com, NvM, Dem, Dcm, Os, MemIf), ARXML debugging, RTE generation troubleshooting, and CAN COM-stack debugging (CanIf -> PduR -> Com). |
Classic + Adaptive |
misra |
Audit C against MISRA C:2025 or write code that is compliant by construction. |
Classic (MISRA C) |
code-review |
Embedded C correctness review (ISR safety, integer overflow, races, stack) plus AUTOSAR naming-convention audit. |
Classic + Adaptive |
requirements |
Elicit, refine, and trace requirements in EARS notation with ASIL attributes. |
Neutral |
iso26262 |
Run a HARA to assign ASIL, or derive Safety Goals and Functional Safety Requirements with FTTI. |
Neutral |
embedded-testing |
Generate MC/DC-covering unit tests and systematic boundary-value test points. |
Classic + Adaptive |
embedded-debugging |
Triage a field problem report or debug a specific fault (HardFault, watchdog, Dem event, stack overflow). |
Classic + Adaptive |
change-and-impact |
Analyze a change request before work begins, or trace direct and indirect ripple effects with a regression scope. |
Classic + Adaptive |
Domain coverage: AUTOSAR Classic (CP) and Adaptive (AP), BSW, MCAL, RTE, AUTOSAR OS, ara::com, ara::exec, SOA, MISRA C, ISO 26262, ASIL, functional safety, ECU, CAN.
Why it exists
Embedded automotive work carries a lot of domain context - AUTOSAR layering, MISRA C rules, ISO 26262 ASIL logic - that general AI tools do not know. Without it you re-explain the same background on every prompt. These skills encode that context once so the AI gives consistent, expert-level answers without the re-explaining.
Usage
The two commands above cover most needs. For finer control:
# Install one skill (use the name shown by --list)
npx skills add ptsilivis/autonomousguy --skill autosar-bsw
# Target a specific agent
npx skills add ptsilivis/autonomousguy -a claude-code
npx skills add ptsilivis/autonomousguy -a copilot
# Non-interactive (CI)
npx skills add ptsilivis/autonomousguy -y
After install, each skill lands in your agent’s native location (e.g. .claude/skills/<name>/SKILL.md). Invoke it the way you invoke any Agent Skill.
Supported agents
Claude Code, GitHub Copilot, Cursor, Codex, Gemini CLI, opencode, and any tool that supports the Agent Skills convention. Path resolution is handled by the skills CLI per agent. Pass -a <agent> to target one explicitly; without it the CLI detects installed agents.
Requirements
- Node.js >=18 (for
npx).
- An AI tool that supports Agent Skills.
Contributing
This is early. Feedback on gaps - missing skills, weak guidance, wrong domain detail - is wanted. Open an issue: https://github.com/ptsilivis/autonomousguy/issues
To work on the repo:
git clone https://github.com/ptsilivis/autonomousguy.git
cd autonomousguy
node bin/validate.js # validate skill structure
Standards and licensing notice
The skills reference two copyrighted standards:
- MISRA C:2025 - © The MISRA Consortium Ltd. Sold per-seat; not redistributable.
- ISO 26262:2018 (all parts) - © ISO. Sold per-part; not redistributable.
This library cites rule and clause identifiers (e.g. “MISRA C:2025 Rule 11.3”, “ISO 26262-3:2018 Table 4”) and paraphrases their intent in its own words. It does not reproduce normative text, rationale, amplification, examples, or full lookup tables. To apply these skills on a real project you must hold a properly licensed copy of each standard. If you spot content that reproduces normative text rather than paraphrasing it, open an issue and it will be rewritten.
License
MIT © Panagiotis Tsilivis
评论