Execute SKILL code on Virtuoso. Use when running SKILL expressions, querying cellview data, listing libraries/cells, or interacting with Virtuoso programmatically.
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Execute SKILL code on Virtuoso. Use when running SKILL expressions, querying cellview data, listing libraries/cells, or interacting with Virtuoso programmatically.
argument-hint
["SKILL expression to run"]
allowed-tools
Bash(vcli *) Bash(virtuoso *)
Execute SKILL Code on Virtuoso
Run SKILL expressions via vcli skill exec and parse results.
Quick reference
# Arithmetic
vcli skill exec"1+2" --format json
# String operations
vcli skill exec'strcat("hello" " " "world")' --format json
# List all libraries
vcli skill exec'foreach(mapcar lib ddGetLibList() lib~>name)' --format json
# List cells in a library
vcli skill exec'let((lib) lib=ddGetObj("myLib") foreach(mapcar c lib~>cells c~>name))' --format json
# Get cell views
vcli skill exec'let((cell) cell=ddGetObj("myLib" "myCell") foreach(mapcar v cell~>views v~>name))' --format json
vcli skill --format json
vcli skill --format json
vcli skill --format json
vcli skill --format json
When multiple Virtuoso instances are running, specify the session explicitly:
vcli session list # find alive session IDs
vcli --session meowu-meow-38371 skill exec'getCurrentTime()'export VB_SESSION=meowu-meow-38371 # or set once for the shell
vcli skill exec'getCurrentTime()'
Connection failure recovery
⚠️ Gotcha: --session <id> bypasses auto-filtering. If Virtuoso restarts, the
session ID changes (new port → new ID). Explicitly specifying a dead session returns
connection_failed: Connection refused (exit 1) even though other sessions are alive.
Recovery pattern (observed signal: exit 1 on skill exec followed by session list):
# 1. Stale session detected
vcli --session meowu-meow-32987 skill exec'getCurrentTime()'# → connection_failed: Connection refused (exit 1)# 2. Purge dead session files and find new IDs
vcli session cleanup
vcli session list
# 3. Retry with the new session ID
vcli --session meowu-meow-38371 skill exec'getCurrentTime()'
Multi-line input — skill eval --stdin
For snippets that span more than one line, or that contain characters painful to
shell-quote (double quotes, parens, dollar signs), pipe the SKILL through stdin
instead of stuffing it on the argv:
# argv form — fine for one-liners
vcli skill exec'1+2' --format json
# stdin form — survives any quoting problem; here-doc friendlycat <<'EOF' | vcli skill eval --stdin --format json
let((x y)
x = 1 + 2
y = x * x
y
)
EOF
Both skill exec and skill eval go through the daemon, but they differ in two
ways that matter:
skill exec
skill eval
Input
argv only (<code> positional)
argv OR --stdin
Multi-statement
daemon let((r) r=<code>) — single form only
wrapped in progn(\n<code>\n) — any number of forms
Trailing ; comment
safe
safe (the \n) before ) terminates the line comment)
Return value
the daemon's r
the value of the last form inside progn
Rule of thumb: use eval --stdin whenever your SKILL has more than one top-level
form, contains full-line comments, or defines a procedure then calls it.
⚠️ Gotcha — skill exec silently drops earlier forms. Inside the daemon's
let(((__vb_r <code>)) ...), f1() f2() parses as "apply the result of f1() to
the arguments of f2()" — only f2() runs. If you see printf(...) succeed but
the side effect you expected from the prior statement is missing, switch to
skill eval --stdin so the input is wrapped in progn(...).
CIW output vs return value
vcli skill exec (and eval) send the SKILL expression to Virtuoso for evaluation
and capture the return value back into the JSON output field. They do not
echo anything to the CIW window unless your expression does it itself with printf.
# Pure return value — CIW stays silent; vcli stdout gets "3"
vcli skill exec'1+2' --format json
# → {"status":"success","output":"3","errors":[],"warnings":[],...}# CIW: (nothing)# Both — printf shows in CIW, the let body's final form is the return value
vcli skill exec'let((v) v=1+2 printf("1+2 = %d\n" v) v)' --format json
# → {"status":"success","output":"3",...}# CIW: 1+2 = 3
When debugging a long expression, the printf(...) + final-form pattern is the
cleanest way to inspect intermediate state without leaving CIW trace breadcrumbs
that pollute later reads.
Attribution
The multi-line SKILL wrapping (progn(\n<code>\n)), --stdin input mode, and
the CIW output vs return-value distinction are adapted from
virtuoso-bridge-lite
(MIT, 2026-06) — specifically cli.py::cli_eval,
examples/01_virtuoso/basic/05_multiline_skill.py, and
examples/01_virtuoso/basic/00_ciw_output_vs_return.py.
Important notes
Use --format json for structured output (auto in pipe mode)
Use --timeout N for long-running operations (default 30s)
SKILL strings use ", escape with \" inside bash single quotes
let blocks work for local variables; Ocean functions (simulator, design, run) must be at top level
View names may not be standard schematic — check with v~>name not v~>viewName
Always wrap slot access in let: let((cv) cv=geGetEditCellView() cv~>cellName) not geGetEditCellView()~>cellName — the latter crashes if no design is open
