Playwright

Playwright CLI Skill

Drive a real browser from the terminal with playwright-cli and the bundled wrapper. Do not pivot to @playwright/test unless explicitly asked for tests.

Prerequisite check (required)

Prefer an existing playwright-cli. The wrapper fails closed rather than installing from the network unless the user authorises package retrieval:

command -v playwright-cli || command -v npx

If both are absent, report the missing capability. Do not install Node or a CLI without authority. For an authorised one-run npx resolution, set:

PLAYWRIGHT_CLI_ALLOW_NPX_INSTALL=1 "$PWCLI" --help

An existing PATH binary is used as installed; the wrapper does not attest its version. When lineage matters, record command -v playwright-cli and playwright-cli --version. Only the authorised npx fallback is package-pinned; --help proves available commands, not version identity.

Skill path (set once)

export AGENTS_HOME="${AGENTS_HOME:-$HOME/.agents}"
export PWCLI="$AGENTS_HOME/skills/playwright/scripts/playwright_cli.sh"

Core workflow

1. Open the page.
2. Snapshot to get stable element refs.
3. Interact using refs from the latest snapshot.
4. Re-snapshot after navigation or significant DOM changes.
5. Capture only assigned artifacts; close the run-owned session.

Minimal loop:

"$PWCLI" open https://example.com
"$PWCLI" snapshot
"$PWCLI" click e3
"$PWCLI" snapshot

References

Open only what you need:

• CLI command reference: references/cli.md
• Practical workflows and troubleshooting: references/workflows.md

Guardrails

• Before submit, send, purchase, delete, account/config mutation, upload or any

other external effect, resolve action-specific authority and target. Opening or inspecting a page does not authorise the effect.

• Never place real secrets in command arguments. Inventory whether the session

is signed in; prefer synthetic accounts/data and an isolated named session.

• Screenshots, snapshots, downloads and traces may contain credentials, personal

data, DOM state, console/network content or tokens. Write only to an assigned run path, declare retention, and remove only proven run-owned artifacts.

• Always snapshot before referencing element ids like e12.
• Re-snapshot when refs seem stale.
• Prefer explicit commands over eval and run-code unless needed.
• When you do not have a fresh snapshot, use placeholder refs like eX and say why; do not bypass refs with run-code.
• Use --headed when a visual check will help.
• Close run-owned sessions and report any session/artifact retained for handoff.
• Default to CLI commands and workflows, not Playwright test specs.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: mblauberg
• Source: mblauberg/provenant
• License: MIT

Install and usage instructions live in the source repository linked above.