agentleFS
RAMA-L7RAMA-L7/vlsi-pd-staff-engineer/CLAUDE.md

vlsi-pd-staff-engineer

CLAUDE.md3 starsChanged 6 months agoMITView on GitHub

Multi-agent framework for RTL-to-GDSII physical design using Yosys + OpenSTA + OpenROAD on Sky130 PDK. This project uses Claude Code's agent/skill/hook system to run a disciplined PD flow with verification gates between every stage. Every PD stage must pass three checks before the next stage starts: 1. QoR reviewer approves metrics (fresh sub-agent audit) 2. LEC verifies equivalence against the previous stage 3. Low power agent checks domain integrity If 3 ECOs don't resolve a failure, STOP and escalate. Do…

The file

We split this CLAUDE.md into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every PD stage must pass three checks before the next stage starts:

    line 31
  2. Distinctive:

    If 3 ECOs don't resolve a failure, STOP and escalate.

    line 38

≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

VLSI Physical Design Staff Engineer

Multi-agent framework for RTL-to-GDSII physical design using Yosys + OpenSTA + OpenROAD on Sky130 PDK.

Quick Start

This project uses Claude Code's agent/skill/hook system to run a disciplined PD flow with verification gates between every stage.

Stage Agents (RTL-to-GDSII pipeline)

  • synthesis-engineer — RTL compilation, multi-corner optimization, QoR targeting
  • floorplan-engineer — die sizing, macro placement, power grid, pin assignment
  • placement-engineer — global/detailed placement, congestion, timing-driven opt
  • cts-engineer — clock tree synthesis, skew balancing, useful skew
  • routing-engineer — global/detail routing, DRC convergence, hold fixing
  • signoff-engineer — multi-corner STA, DRC, LVS, IR drop, EM, GDSII

Cross-Cutting Agents (run at every stage boundary)

  • lec-verifier — formal equivalence checking at every stage boundary
  • low-power-verifier — UPF/CPF domain verification (different checks per stage)

Reactive Agent (dispatched on failure)

  • forensic-eco-debugger — four-phase investigation, five-level trace, three-fix rule

Stage Gate Rules

Every PD stage must pass three checks before the next stage starts:

  1. QoR reviewer approves metrics (fresh sub-agent audit)
  2. LEC verifies equivalence against the previous stage
  3. Low power agent checks domain integrity

Three-Fix Rule

If 3 ECOs don't resolve a failure, STOP and escalate. Do not attempt ECO #4 without human approval and five-level root cause trace.

Tool Environment

Session-start hook auto-detects: Yosys, OpenSTA, OpenROAD, PDK path, library files, and QoR targets from synthesis_targets.json.

Project Structure

vlsi-pd-staff-engineer/
├── CLAUDE.md                              # this file
├── synthesis_targets.json                 # QoR targets + corner definitions
├── agents/                                # team roster (9 agents)
│   ├── synthesis-engineer.md              #   RTL → gate netlist
│   ├── floorplan-engineer.md              #   die + macros + power grid
│   ├── placement-engineer.md              #   cell placement + legalization
│   ├── cts-engineer.md                    #   clock tree synthesis
│   ├── routing-engineer.md                #   metal wires + DRC clean
│   ├── signoff-engineer.md                #   final verification + GDSII
│   ├── lec-verifier.md                    #   equivalence at every boundary
│   ├── low-power-verifier.md              #   domain checks at every stage
│   └── forensic-eco-debugger.md           #   reactive failure investigation
├── skills/                                # process library (10 skills)
│   ├── using-pd-staff-engineer/SKILL.md   #   culture deck (loaded every session)
│   ├── synthesis/SKILL.md                 #   641 lines — optimization strategies
│   ├── synthesis/qor-reviewer-prompt.md   #   QoR audit sub-agent
│   ├── floorplan/SKILL.md                 #   macro placement, PG, pins
│   ├── placement/SKILL.md                 #   congestion, timing-driven
│   ├── cts/SKILL.md                       #   skew, buffers, Bug #3 territory
│   ├── routing/SKILL.md                   #   DRC convergence, hold fixing
│   ├── signoff/SKILL.md                   #   MMMC STA, tapeout checklist
│   ├── lec/SKILL.md                       #   9 checkpoint protocol
│   ├── lec/lec-reviewer-prompt.md         #   LEC audit sub-agent
│   ├── low-power/SKILL.md                 #   cross-cutting domain checks
│   └── forensic-debugging/SKILL.md        #   four-phase ECO investigation
├── hooks/                                 # management layer
│   ├── hooks.json                         #   session-start trigger
│   ├── run-hook.cmd                       #   cross-platform runner
│   └── session-start                      #   env detection + handbook load
└── commands/                              # custom CLI commands (future)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.