cgagentharness-verify

Verify CG-agent-harness backend, desktop, and model behavior with isolated fixtures and evidence.

Updated Sep 7, 2026
One-click install
npx skills add https://github.com/cgfixit/CG-agent-harness --skill cgagentharness-verify-cgfixit
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: cgagentharness-verify
Source: https://github.com/cgfixit/CG-agent-harness/tree/main/.codex/skills/cgagentharness-verify
Command: npx skills add https://github.com/cgfixit/CG-agent-harness --skill cgagentharness-verify-cgfixit

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Verifying a Rust-based local agent harness across backend, native desktop, and model runtime surfaces is error-prone: tests can touch real operator homes, sandbox errors get mistaken for regressions, and smoke tests may accidentally enable cloud paths. This Skill provides a disciplined verification workflow that ties every result to the tested source SHA, binary, and architecture. ## Core Features & Use Cases - Contract-driven check selection: Maps changed areas to the right checks, such as fmt, Clippy with warnings denied, cargo-deny, policy/approval tests, and macOS Seatbelt-dependent native execution tests. - Isolated live acceptance: Starts the known binary in an owned temporary home on a unique port, issues one harmless local chat, and verifies clean shutdown without touching operator data. - Evidence-based reporting: Records commands, exit statuses, tested SHA/architecture, artifact hashes, skips, and remaining limits while excluding credentials and home data. - Use Case: Before a release, run smoke tests and regression checks against the packaged desktop backend, confirm the exact configured model tag, and produce an acceptance report linked to the tested binary. ## Quick Start Verify the current CG-agent-harness checkout by running the backend quality gates and a live acceptance smoke test in a temporary home, then report the tested SHA and results.

Frequently Asked Questions about cgagentharness-verify

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run smoke tests for the CG-agent-harness backend?▼

Run the root-toolchain fmt, Clippy with warnings denied, all-target tests, and cargo-deny as documented in AGENTS.md. For live acceptance, start the known binary in an owned temporary home on a unique port, issue one harmless local chat, and verify clean stop.

How to verify the macOS desktop app before release?▼

Package the backend before compiling the shell, then run desktop-toolchain fmt, Clippy/tests, and scripts/test-desktop-backend.py against the packaged backend. Follow docs/DESKTOP_ACCEPTANCE.md for outstanding UI checks and record only interactions actually observed.

Why do native execution tests fail with sandbox permission errors?▼

Tests like tests/macos_cargo.rs and process lifecycle tests require real macOS Seatbelt. Sandbox permission errors come from the environment, not the application, so they are not application regressions and should be diagnosed as such.

Can fake model tests prove live model availability?▼

No. Fake model and provider tests only prove protocol behavior, not availability or live quality. For model verification, inspect the actual inventory and exact configured model tag, and never enable cloud fallback during a smoke test.

What should a verification report include?▼

Report commands, exit statuses, tested source SHA and architecture, artifact hashes, skips, and remaining limits. Never include credentials or operator home data, and close only processes started by the verification itself.