Loading...
Loading...
Found 3,749 Skills
Runs ML experiments reproducibly — single runs or autonomous BFS batches. Single mode: isolated venv, time-budgeted, failure-handled, logs to RESEARCH.md. BFS mode (opt-in): designs N hypotheses, runs each for a fixed budget, compares via a single verifiable metric, keeps improvements and git-resets failures — fully autonomous until done. Respects the RESEARCH.md supervision policy for notifications, approvals, and stop limits. Trigger phrases: "run experiment", "train model", "explore design space", "find best config", "autoresearch".
Reply to every PR review comment — from GitHub Copilot, other review bots, and human reviewers — with code fixes or reasoned push-backs. Fetches all comments via the GitHub API, triages each one, fixes code, commits, and posts an inline reply to every comment. Use whenever the user wants to address PR feedback, respond to review comments, handle Copilot's review, clear out reviewer comments, or says 'address the review', 'reply to the comments', 'handle the PR feedback' — even if no specific reviewer is named. Not complete until every comment has a response posted.
Automatically discover CI/CD and automation skills when working with GitHub Actions, Jenkins, GitLab CI, pipelines, continuous integration, continuous deployment, or automated testing. Activates for CI/CD development tasks.
Use when researching another Git repository
Automated factory that converts GitHub repositories into standardized AI Skills. This tool is used when users provide a GitHub URL and want to "package", "wrap", or "create a Skill". It supports automatic retrieval of repository metadata, generation of standard directory structures, and injection of extended metadata required for lifecycle management.
Use when babysitting a PR/MR until CI is green and every valid reviewer feedback is addressed — supports GitHub PR (gh) and GitLab MR (glab), triages comments into Valid / Discuss / Out-of-scope, addresses valid items with small commits and inline thread replies, escalates invisible findings (SonarQube/Snyk dashboards) and 3-round bot deadlocks, reports ready-to-merge (never auto-merges). Triggers — '監看 PR', 'babysit PR/MR', 'PR 顧到 merge', 'address review feedback', 'wait until CI green', '把 PR 顧到綠'. NOT for writing PR descriptions, NOT for diff code review (use pr-review), NOT for actually merging the PR (user does that).
Autonomous NeMo-RL research agent workflow for directed hypothesis testing and open-ended discovery. Guides agents through the full experiment lifecycle: understanding recipes and environments, wiring RL or NeMo-gym runs, launching reproducible baselines and iterations, analyzing results, preserving human oversight, and using git plus TSV logs as the research ledger. Do NOT use for: bug fixes, code review, documentation, refactoring, dependency updates, or single-file changes.
End-to-end GitHub issue fix workflow using gh, local code changes, builds/tests, and git push. Use when asked to take an issue number, inspect the issue via gh, implement a fix, run XcodeBuildMCP builds/tests, commit with a closing message, and push.
Use Git worktrees for isolated work environments. Creates separate working directories for parallel development on different branches.
Manages deferred/blocked work items as GitHub Issues with labels and blocker tracking
Conventional Commits v1.0.0 standards for git messages. Use when (1) creating git commits, (2) writing or drafting commit messages, (3) reviewing commit message format, (4) explaining commit conventions, or (5) validating commit message compliance.
After the task execution is completed, prompt the user to open a new Agent to review the uncommitted git code. Athletes should not act as referees; proceed with the wrap-up only after the review is approved.