description: Scaffold a new canonical MBG skill under .ai/skills using the project authoring standard, then resync all platform adapters.

Create a new canonical skill for the Microelectronic Block Generator project.

Requested input:

$ARGUMENTS

Interpret the first argument as the skill name. Treat the remaining arguments as the requested purpose or capability.

Requirements

Load the mbg-extension-authoring skill.
Require a skill name beginning with mbg-.
Require lowercase alphanumeric kebab-case.
Reject names containing spaces, underscores, consecutive hyphens, or leading/trailing hyphens.
Create the skill at:
```
.ai/skills/<skill-name>/SKILL.md
```
Do not create it directly under .opencode/skills/, .claude/skills/, or plugins/mbg-analog/skills/ — those are generated output; a hand-edit there is discarded the next time the sync script runs.
Ensure the directory name exactly matches the YAML frontmatter name.
Do not overwrite an existing skill without explicit approval.
Give the skill one clear responsibility.
Set the correct project owner (huda, ahmad, or jabir) and list real capability ids that either already exist in .ai/manifest.json's capabilities section or are being added alongside this skill.
Do not invent tools, APIs, modules, or verification results — verify every referenced entry point against the actual source before writing it into the skill.

PDK Body Constraint

MOSFET body: pfet_03v3 -> VDD ONLY, nfet_03v3 -> VSS ONLY. If the new skill generates or processes SPICE netlists, it must enforce this rule.

Required Frontmatter

Flat key: value lines only; list fields written inline. The sync script's parser is minimal (no PyYAML in this repo's Python environments) — nested maps, block lists, and multi-line folded scalars are not supported.

---
name: mbg-example-skill
description: 2-4 sentences — what it does, when to use it, when not to.
class: READ-ONLY | GENERATING | MUTATING | DESTRUCTIVE
owner: huda | ahmad | jabir
capabilities: [cap_id]
platforms: [opencode, claude, codex]
---

Replace the example name, description, class, owner, and capabilities with values appropriate to the requested skill.

Required Sections

The generated skill body must include, per docs/opencode/AUTHORING_GUIDE.md's section layout: ## Purpose, ## When to Use, ## When Not to Use, ## Required Inputs, ## Preconditions, ## Workflow, ## Outputs, ## Failure Modes.

Keep the body platform-neutral — do not mention a specific platform by name unless the skill is genuinely about authoring for that one platform.

Safety Requirements

Never read or display .env.
Never include API keys or credentials.
Never include personal absolute filesystem paths.
Never use unlimited retry or refinement loops.
Never claim simulation or verification success without evidence.
Never modify another team member's implementation outside the assigned scope.
Never stage, commit, or push automatically.

Validation

After creating the skill:

Confirm that the directory and frontmatter name match.
Confirm the frontmatter is flat (no nested YAML).
Confirm every referenced module/function actually exists in the repo.
Report every validation error and warning found.
Do not claim completion if validation fails.
Run python3 scripts/sync_agent_tools.py to regenerate the OpenCode, Claude, and Codex adapters from the new .ai/skills/ entry. If the skill introduces a new capability, coordinate with whoever owns .ai/manifest.json to add the corresponding entry there — this workflow does not hand-edit the manifest itself.

Output Format

Skill:
Owner:
Purpose:
File created:
Manifest capability entry needed:
Sync command run: python3 scripts/sync_agent_tools.py
Validation status:
Errors:
Warnings:
Known limitations:
Git status:

If no skill name is supplied, stop and request a valid mbg- skill name.

Créer un nouveau skill MBG

Recommandé pour