Andrew Sun
Community@victoriacity · Minneapolis, MN
Head of Applied Research @ Meshy AI, Machine learning, molecular simulation, computer graphics, visual design
Agent Skills by Andrew Sun
Showing 25 vetted skills indexed across 1 GitHub repositories.
slack-diagnosis
Diagnose Slack bot issues by correlating conversation threads, system logs, and application code.
design
Design rigorous empirical studies and experiment protocols from research questions.
postmortem
Analyze flawed AI agent outputs to identify reasoning failures and propose preventative measures.
refresh-skills
Audit and update SKILL.md files to align with the current codebase.
coordinator
Execute operational commands and manage workflows for Slack-based AI assistants.
develop
Develop and refactor Akari infrastructure code using test-driven development.
gravity
Scan documentation and logs to identify recurring manual tasks for formalization.
orient
Assess agent state, prioritize tasks, and identify mission gaps.
compound-simple
Analyze Git diffs and generate tasks for project convention updates.
diagnose
Analyze experimental result files to identify error patterns and root-cause hypotheses.
feedback
Process human feedback to implement fixes and record learnings.
orient-simple
Select the highest-scored unblocked task from TASKS.md for an AI agent.
project
Create research projects via agent-driven gap analysis or interactive scaffolding.
publish
Automate research paper publication pipelines for venue or arXiv submission.
architecture
Diagnose architectural issues like duplication and coupling in codebases.
review
Validate experiment metrics and findings through two-stage review.
lit-review
Searches, verifies, and organizes literature into structured notes and tasks.
compound
Embed session learnings into system conventions, skills, and patterns.
report
Generate formatted markdown reports and chart images from Akari data sources.
critique
Review project plans, designs, and findings against 10 failure dimensions.
self-audit
Audit recent Git commits for CLAUDE.md convention compliance.
simplify
Audit plans or implementations against a stated mission to identify unnecessary complexity.
horizon-scan
Scan external sources for Generative AI developments and record updates.
audit-references
Fetch URLs and compare titles and authors to verify cited references.