Run local Pi coding-agent workflows with sandbox and session-sharing guardrails
Use Pi when an operator needs a local coding-agent CLI with explicit sandboxing, provider setup, and reviewable session artifacts.
npx skills add agentskillexchange/skills --skill run-local-pi-coding-agent-workflows-with-sandbox-and-session-sharing-guardrails
Use Pi when an agent or operator needs to run a local interactive coding-agent workflow from a terminal while keeping provider configuration, sandboxing, and session-sharing behavior explicit. The workflow is to install the Pi coding-agent CLI, run it against a bounded repository task, optionally containerize or sandbox execution, and publish or preserve OSS session traces for later review. Invoke this instead of using Pi normally when the job is a repeatable operator runbook for supervised coding-agent execution, not generic LLM API use or terminal UI development. The boundary is local coding-agent operation with sandbox/session controls, not the full Pi monorepo or every library in the project.
What this skill actually does
Inputs and prerequisites: Pi coding-agent CLI, configured LLM provider credentials, target repository, optional Docker/OpenShell/Gondolin sandbox.
Setup notes: Install the @earendil-works/pi-coding-agent package from npm, configure the intended model provider, run Pi from the target repository, and apply the documented containerization or sandbox pattern before allowing broad filesystem or shell access.
Source and verification boundary: use https://pi.dev/docs/latest as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Custom Agents workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.