Dispatch, monitor, and evaluate connections-checker E2E runs against main baselines on the AllSpice Hub.
npx skills add https://github.com/shrik450/dotfiles --skill run-e2e-shrik450
Please help me install this Agent Skill. Skill: run-e2e Source: https://github.com/shrik450/dotfiles/tree/main/dot_claude/skills/run-e2e Command: npx skills add https://github.com/shrik450/dotfiles --skill run-e2e-shrik450
e2e_manual.yml with a full 40-character SHA, then poll run status, design reviews, and comments via the hub.py helper script. - Summarize results: Use summarize.py to condense rubric judge comments and metrics-index bands into compact per-run blocks for comparison against main baselines. - Pull artifacts and traces: Download eval_output artifacts with pull_artifact.py to inspect rendered page input, model reasoning groups, and verifier confidence reports. - Use Case: After pushing a branch that changes how KiCad schematics are rendered, dispatch an E2E round, wait 15-25 minutes, then compare caught rubric items and token usage against four recent main baselines to confirm the change engaged without regressions. ## Quick Start Ask the assistant to run an E2E round for the current connections-checker branch and report how it compares to recent main baselines.High-intent search queries and answers about installing and using this skill.
Push the ref to GitHub, then dispatch the e2e_manual.yml workflow with the full 40-character SHA using gh workflow run. Poll run status with the hub.py tasks subcommand every 90 to 120 seconds until the design reviews complete.
Find recent main runs with hub.py drs, taking four including the branch's merge base, then run summarize.py on each design review. Judge caught rubric items, false-positive buckets, and token usage against the baseline spread rather than a single run.
An abbreviated commit SHA causes every run to fail in roughly 30 seconds because the injected workflow cannot resolve the action ref. Re-dispatch using the full 40-character SHA from git rev-parse HEAD.
No. The Gitea 1.24 Hub exposes no job-log endpoint, and the web log route requires a session cookie rather than a token. When a run fails early, ask the user to paste the failing step from the run page.
Download the eval_output artifact with pull_artifact.py and read debug_dump pages input.txt, which contains the exact rendered page text fed to the model. Diffing it against a baseline run shows precisely what the change introduced.
googleRun unit tests and lint checks before pushing changes to GitHub.
jeremybrasherFilters false positives from review findings using three self-challenge rounds and an arbiter verdict.
FilenCloudDiensteDecompose complex requests and delegate sub-tasks to subagents.
RaiderIOGenerates a Toggl-compatible CSV timesheet from git commits and Claude session history.