rat-init-project
Initialize RTL project workspace: directory structure, coding rules, phase guides.
Skill by babyworm
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/babyworm-rat-init-project
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Shell scripts
Network hosts it talks to 2
codeload.github.com· found in install commandgithub.com· found in code
Files it touches
~/.claude/skills/rat-init-project
What's inside
- Skill name
rat-init-project- Description
- Initialize RTL project workspace: directory structure, coding rules, phase guides. Run once per project. Triggers: 'init project', 'new project'.
- License
- MIT
- Instructions
- 323 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
- Instructions read files in your home folder (~/.ssh, ~/.aws…)
Agent Skills spec issues
- files: referenced but missing: scripts/run_sim.sh
Files (24)
- SKILL.md
- generate_config.sh · scripts/
- inject_claude_md.sh · scripts/
- install_project_templates.sh · scripts/
- Makefile · templates/
- claude-md-rat-section.md · templates/
- cocotb-makefile · templates/
- filelist.f · templates/
- docs-guide.md · templates/guides/
- refc-guide.md · templates/guides/
- reviews-guide.md · templates/guides/
- rtl-guide.md · templates/guides/
- sim-guide.md · templates/guides/
- syn-guide.md · templates/guides/
- tool-runner.sh · templates/lib/
- phase6-pdf-makefile · templates/
- rat_config.json · templates/
- rtl-coding-conventions.md · templates/rules/
- rtl-verification-gate.md · templates/rules/
- run_cdc.sh · templates/
- run_conformal.sh · templates/
- run_formality.sh · templates/
- run_lint.sh · templates/
- run_syn.sh · templates/
SKILL.md (first lines)
<Purpose>
Initialize a new RTL design project with the standard directory structure, coding convention
rules, phase guides, and template files required by the 6-Phase pipeline.
This is a per-project operation — run once per new project workspace.
For EDA tool installation and environment verification, use `/rtl-agent-team:rat-setup` instead.
</Purpose>
<Use_When>
- Starting a new RTL/FPGA/ASIC design project
- First time using this plugin in a new workspace
- User says "init project", "initialize project", "new project", "project init"
- Agent Step 0 detects missing project structure (.claude/rules/ not found)
</Use_When>
<Do_Not_Use_When>
- Project directories already exist and rules are deployed (check `.claude/rules/rtl-coding-conventions.md`)
- Only need to verify/install EDA tools (use `/rtl-agent-team:rat-setup`)
- Designing architecture or writing RTL (use p2-arch-design or rtl-p4-implement)
</Do_Not_Use_When>
<Why_This_Exists>
The 6-Phase pipeline expects a standard directory layout (rtl/, refc/, bfm/, sim/, lint/, syn/, etc.)
and coding convention rules deployed to `.claude/rules/`. Without proper project initialization,
agents fail with missing directory errors and coding conventions are not enforced.
This skill ensures the project workspace is ready before design work begins.
</Why_This_Exists>
<Execution_Policy>
- Do not overwrite existing project templates, guides, rules, or directories
- Managed exceptions: update only the RAT block in project `CLAUDE.md`, refresh
`rat_config.json` tool detection, and regenerate `config.mk`
- Preserve usable user tool path overrides; refresh stale or unusable paths from
the detected environment
- Generate a project initialization report at the end
- Quick EDA tool status check (informational only — does NOT install tools)
- If required tools are missing, recommend running `/rtl-agent-team:rat-setup` for installation
</Execution_Policy>
<Steps>
1. **Check plugin setup** (advisory, not a hard gate):
Check `${CLAUDE_PLUGIN_DATA:-$HOME/.config/rtl-agent-team}/.setup-complete` for `/rtl-agent-team:rat-setup` completion.
If marker is missing, advise: "Run `/rtl-agent-team:rat-setup` first to verify EDA tools and configure the plugin environment."
Proceed anyway — project structure can be created without tools, but warn about missing EDA environment.
1a. **Check project root**: Verify current directory is suitable (has .git or is empty).
2. **Create directory structure** (skip existing):
```
specs/ # Input specifications and datasheets
refc/ # C reference model (DPI-C compatible)
include/ # Common ref model headers
build/ # Build output (.so for DPI-C)
bfm/ # Bus Functional Models
include/ # Common BFM headers
rtl/ # Synthesizable SystemVerilog source
common/ # Shared utility modules (ICG, synchronizer, CDC primitives)
include/ # Common defines, packages
top/ # Top-level module instantiation
sim/ # Simulation & testbenches
top/ # Tier 4: integration tests
formal/ # SVA formal verification (.sby configs)
lint/ # Lint flow
scripts/ # Lint scripts (run_lint.sh)
reports/ # Per-module lint results
syn/ # Synthesis flow
scripts/ # Synthesis scripts (run_syn.sh)
reports/ # Per-module synthesis results
docs/ # Design documentation
phase-1-research/ # Phase 1 artifacts
phase-2-architecture/ # Phase 2 artifacts
phase-3-uarch/ # Phase 3 artifacts
phase-4-rtl/ # Phase 4 artifacts
phase-5-verify/ # Phase 5 artifacts
decisions/ # Architecture Decision Records (ADR)
reviews/ # Phase gate review reports (Markdown)
phase-1-research/
phase-2-architecture/
phase-3-uarch/
phase-4-rtl/
phase-5-verify/
phase-6-review/ # Phase 6 review deliverables
.rat/
state/ # Plugin state files (auto-managed)
```
Note: Per-module subdirectories under `refc/`, `bfm/`, `rtl/`, `sim/` are created
during Phase 2 (architecture) when module decomposition is decided.
2a. **Inject project CLAUDE.md** (RAT-managed section):
Run the injection script to create or update the RAT-managed section in the project root CLAUDE.md.
User content outside `<!-- RAT:START -->` / `<!-- RAT:END -->` tags is never modified.
```bash
bash "${CLAUDE_PLUGIN_ROOT}/skills/rat-init-project/scripts/inject_claude_md.sh" .
```
Idempotent — safe to re-run on every `rat-init-project` invocation.
2b. **Generate project config** (`rat_config.json`):
Run the config generator to detect available EDA tools and create the project configuration file.
If `rat_config.json` already exists, tool availability and paths are refreshed while
user-owned configuration is preserved. A usable explicit path override remains unchanged;
a stale or unusable path is replaced when the tool is found in the current environment.
```bash
bash "${CLAUDE_PLUGIN_ROOT}/skills/rat-init-project/scripts/generate_config.sh" . "$(basename "$(pwd)")"
```
The config file stores:
- `tools`: 24 EDA tools across 8 categories (simulators, synthesis, lint, formal, equivalence, cdc, debug, coverage). Each tool has `detected`, `path`, and `env_source` fields. Set `env_source` per-tool for tools needing setup (e.g., `"env_source": "source /tools/synopsys/vcs/setup.sh"`). Each `env_source` is trusted project input and is evaluated by Bash on every generator run; do not populate it from untrusted data.
- `preferences`: preferred tool per category (auto-set to first detected commercial tool, user-overridable)
- `technology`: target library path, SRAM lib, NAND2 cell pattern (auto-extracted from liberty if set)
- `coverage`: targets (line≥90%, toggle≥80%, FSM≥70%, branch≥80%, functional≥95%), seeds, fail rate
- `waivers`: custom paths for lint/CDC waiver files
Users should edit per-tool `env_source`, `technology.liberty`, and `waivers` after generation.
`config.mk` is plugin-managed output and is regenerated from the refreshed configuration.
2c. **Deploy rules** (skip if already deployed globally via `/rtl-agent-team:rat-setup`):
For each rule file, check `~/.claude/rules/` first. If the same file exists globally,
skip the local copy to avoid duplicate injection. Only deploy locally if neither exists.
Diagram rules are injected into `~/.claude/CLAUDE.md` by `/rtl-agent-team:rat-setup` via `<markdown_diagram_rule>` tag.
If the tag is missing (`/rtl-agent-team:rat-setup` not run), write an inline fallback file.
```bash
mkdir -p .claude/rulesFrom babyworm/rtl-agent-team, MIT.
More from babyworm/rtl-agent-team
- 0.628score
rtl-bug-repro
babyworm
Reproduce an RTL bug with a minimal TB and isolate root cause via waveform diff — 'reproduce this bug', 'minimal repro TB', regression failure isolation.
SkillBFreshness: B★ 61Claude Code - 0.628score
rtl-lint-check
babyworm
Lint RTL with Verilator, Verible, and slang; report-only quick check for 'run lint', pre-commit, or phase-gate verification (no fixing).
SkillBFreshness: B★ 61Claude Code +9 - 0.628score
rtl-document
babyworm
Generate Markdown docs for an RTL module (port/param/FSM tables, synth summary) — 'document this module', 'port table', 'refresh RTL docs'.
SkillBFreshness: B★ 61Claude Code - 0.628score
ref-model
babyworm
Builds Phase 2 golden C reference model with bitexact conformance and bandwidth estimation.
SkillBFreshness: B★ 61Claude Code - 0.628score
rtl-ipxact-gen
babyworm
Generate IEEE 1685-2014 IP-XACT XML descriptor from RTL source — 'generate IP-XACT', IP handoff, or SoC/EDA integration needing component XML.
SkillBFreshness: B★ 61Claude Code - 0.628score
codec-conformance-eval
babyworm
Evaluate C ref-model decoder against JVET/JCTVC conformance bitstreams (bitexact/MD5) — 'decoder conformance'; C-level, not RTL (rtl-conformance-test).
SkillBFreshness: B★ 61Claude Code +9 - 0.628score
codec-rd-eval
babyworm
Encoder RD evaluation of C ref model: BD-PSNR/BD-rate (VCEG-M33) across sequences/QPs — 'RD eval', 'BD-rate', codec algorithm quality comparison.
SkillBFreshness: B★ 61Claude Code +9 - 0.628score
rtl-conformance-test
babyworm
Final codec standards compliance gate: bitexact RTL bitstream check against JM/HM reference.
SkillBFreshness: B★ 61Claude Code - 0.623score
domain-consult
babyworm
Route domain questions to the best expert agent: codec standards, video/signal processing, fixed-point math, HW protocols.
SkillBFreshness: B★ 61Claude Code +9 - 0.623score
rtl-p4-block-parallel
babyworm
Phase 4 RTL in isolated git worktrees per codec block, team-coordinated with upstream-first merge.
SkillBFreshness: B★ 61Claude Code - 0.620score
cdc-tool-profiles
babyworm
Internal reference: cdc tool profiles (agent-loaded; do not invoke).
SkillBFreshness: B★ 61Claude Code +9 - 0.620score
rat-auto-design-policy
babyworm
Internal reference: rat auto design policy (agent-loaded; do not invoke).
SkillBFreshness: B★ 61Claude Code +9
Signals
- Stars
- 61
- Forks
- 16
- Open issues
- 0
- Last commit (this folder)
- Aug 23, 2026 (2 months ago)
- Last release
- Jul 17, 2026 (3 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/babyworm/rtl-agent-team
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from babyworm
- 0.628score
rtl-bug-repro
babyworm
Reproduce an RTL bug with a minimal TB and isolate root cause via waveform diff — 'reproduce this bug', 'minimal repro TB', regression failure isolation.
SkillBFreshness: B★ 61Claude Code - 0.628score
codec-rd-eval
babyworm
Encoder RD evaluation of C ref model: BD-PSNR/BD-rate (VCEG-M33) across sequences/QPs — 'RD eval', 'BD-rate', codec algorithm quality comparison.
SkillBFreshness: B★ 61Claude Code +9 - 0.628score
rtl-conformance-test
babyworm
Final codec standards compliance gate: bitexact RTL bitstream check against JM/HM reference.
SkillBFreshness: B★ 61Claude Code - 0.628score
ref-model
babyworm
Builds Phase 2 golden C reference model with bitexact conformance and bandwidth estimation.
SkillBFreshness: B★ 61Claude Code - 0.628score
codec-conformance-eval
babyworm
Evaluate C ref-model decoder against JVET/JCTVC conformance bitstreams (bitexact/MD5) — 'decoder conformance'; C-level, not RTL (rtl-conformance-test).
SkillBFreshness: B★ 61Claude Code +9 - 0.628score
rtl-document
babyworm
Generate Markdown docs for an RTL module (port/param/FSM tables, synth summary) — 'document this module', 'port table', 'refresh RTL docs'.
SkillBFreshness: B★ 61Claude Code
Cite this page
rat-init-project skill. Agent Mart. https://agentmart.dev/skills/babyworm-rat-init-project (last updated 2026-10-24)