omp packages

@diegopetrucci/pi-agent-workflow-audit0.1.14

A pi extension that runs an isolated repo workflow audit and returns only the final report to the main session.

GitHub Actions Published 7 days ago MIT ~500 downloads/mo

Untested on omp

This is a pi package. No one has reported whether it works on omp yet. If you try it, add your result below.

Did it work for you?

Result

One report per person. A new report replaces your old one.

Reports 0

No reports yet.

Readme

agent-workflow-audit

Runs an isolated repo workflow audit from /agent-workflow-audit.

The extension ports the agent-workflow-audit skill into a pi slash command. The audit subagent reads the current repository's agent-facing instructions, manifests, and obvious workflow files, tries or plans the documented setup/build/lint/test/run flow, and reports where instructions or scripts waste agent time.

Why an extension instead of only a skill?

A normal skill teaches the main agent how to run the audit, so all discovery, command output, failures, and retries enter the main session context. This extension runs that noisy work in an isolated in-memory child session and sends only the final distilled report back to the main session.

The final custom message is visible to the main agent and user. Raw child tool transcripts are not added to the main session context.

Install

Standalone npm package

pi install npm:@diegopetrucci/pi-agent-workflow-audit

Collection package

pi install npm:@diegopetrucci/pi-extensions

GitHub package

pi install git:github.com/diegopetrucci/pi-extensions

Then reload pi:

/reload

Usage

/agent-workflow-audit

The command asks for confirmation before execution mode because documented project commands may create dependencies, build artifacts, caches, or other local side effects.

Options:

/agent-workflow-audit --plan-only
/agent-workflow-audit --yes
/agent-workflow-audit focus on release/check/publish docs
  • --plan-only reads docs and manifests, then reports the workflow it would try without running project commands.
  • --yes skips the interactive confirmation.
  • extra words become focus notes for the audit subagent.

Behavior

  • Runs a child createAgentSession with an in-memory session manager.
  • Disables inherited extensions, skills, prompt templates, themes, and context files in the child.
  • Gives the child read, grep, find, ls, and guarded bash tools, plus any configured MCP tools (run without confirmation; may have side effects; requires host Pi >=1.0.4; older hosts degrade to no MCP, no error); the global mcp.json always applies; a project .pi/mcp.json is loaded only when the host session has already trusted the project.
  • Returns a single final agent-workflow-audit custom message to the parent session.
  • Includes a short run-boundary footer noting that intermediate command output, errors, retries, and search path stayed in the child session.

Safety notes

The audit intentionally exercises documented project commands. The runtime guard is conservative: it blocks obvious destructive, publishing, deploy, mutating Git/GitHub, credential-dumping, network shell-installer, path-changing, and source-fixing commands. It may over-block valid workflow commands and report them as manual/friction instead. Project scripts can still create local artifacts, so run it from a checkout where local build/test side effects are acceptable.