gangj277

Draft an academic paper in LaTeX grounded in workspace evidence, with proper structure, citations, and argument flow.

Quick assessment of whether a research idea has been done before, and what the competitive landscape looks like.

Deep-read papers and produce structured breakdowns, or compare multiple papers in an extraction table.

Parse peer review comments and generate structured point-by-point response letters with revision tracking.

Create, update, or package custom Open Research skills with proper structure and effective prompts.

Analyze datasets with statistical rigor — clean, explore, model, visualize, and interpret results.

Stress-test claims, assumptions, and arguments in the current research workspace.

Weigh conflicting evidence and assess which claims are best supported.

Verify whether a candidate coding-agent skill helps in a specific repository on a specific task slice with acceptable risk. Use when an agent needs to compare baseline versus skill-enabled runs, score repo compatibility, inspect instruction and script risk, and leave behind auditable verification artifacts in `.skillproof/`.

Universal repo takeover skill that guarantees safe results in any repository.

TypeScript and React guidance for narrow repository-local refactors. Use when a coding agent needs help reading local code, local tests, and existing conventions before making a small change.

Repository bootstrap guidance for Python and JavaScript projects that may need lightweight helper scripts.

Python and FastAPI backend skill for service endpoints, schemas, and API bugfixes.

TypeScript, Next.js, and React frontend skill for repository-local web-app changes and UI regressions.

Manual red-teaming skill for authorized AI services and agentic repos. Use when you need to inspect a local repo, dev environment, or staging agent for jailbreaks, indirect prompt injection, codebase poisoning, tool or MCP poisoning, memory or RAG poisoning, approval confusion, data exfiltration paths, or concealed malicious actions, and leave behind reproducible findings, regressions, and hardening guidance.

Manual red-teaming skill for authorized AI services and agentic repos. Use when you need to inspect a local repo, dev environment, or staging agent for jailbreaks, indirect prompt injection, codebase poisoning, tool or MCP poisoning, memory or RAG poisoning, approval confusion, data exfiltration paths, or concealed malicious actions, and leave behind reproducible findings, regressions, and hardening guidance.

Grounded forensic evaluation engineering for LLM product repositories. Use when a coding agent needs to inspect an existing codebase, infer the workflow behind a prompt feature, structured output flow, RAG pipeline, tool-calling path, or simple multi-stage chain, then design repo-native evals, capture exact traces, attribute behaviors to prompt or context components, and propose evidence-backed prompt or workflow improvements.

Grounded forensic evaluation engineering for LLM product repositories. Use when a coding agent needs to inspect an existing codebase, infer the workflow behind a prompt feature, structured output flow, RAG pipeline, tool-calling path, or simple multi-stage chain, then design repo-native evals, capture exact traces, attribute behaviors to prompt or context components, and propose evidence-backed prompt or workflow improvements.

Unified Threads publishing and reply automation with Playwright. Use when Codex needs to research a Threads angle, choose between a standalone post/thread, an image-led card post, or a conversation-first reply campaign, draft native Threads copy, discover candidate posts to reply to, validate auth and the active account, run dry-runs or capped live publish/reply sessions, and verify the live result. Triggers on requests to automate Threads posting, create Threads content, publish on Threads, plan or execute a Threads reply campaign, find posts to reply to on Threads, or coordinate both posting and replies as one workflow.