Truong Nguyen
Community@nvtruongops
PI-Guard capstone project delivering prompt-injection and jailbreak detection guardrails, ML classifier training, evaluation metrics, and FPT University thesis documentation.
Agent Skills by Truong Nguyen
Showing 13 vetted skills indexed across 1 GitHub repositories.
nvtruongopsteam-git-sync-and-merge
Enforces workspace-isolated Git branching and leader-only merge workflows for team repositories.
nvtruongopsml-classifier-training
Trains TF-IDF baselines and fine-tunes DeBERTa-v3 transformers for prompt injection detection.
nvtruongopscapstone-thesis-and-defense
Guides writing FPT IAP491 capstone thesis chapters and defense slides for PI-Guard.
nvtruongopsreview1-threat-model-and-defense
Prepares Review 1 thesis chapters on prompt injection threat modeling and guardrail literature review.
nvtruongopsdocs-portal-sync-and-deploy
Synchronizes research documentation into a MkDocs portal and deploys it to GitHub Pages.
nvtruongopsteam-commit-and-workspace-audit
Audits git commits, staging areas, and pull requests for workspace boundary violations.
nvtruongopsguardrail-dataset-engineering
Curates, deduplicates, and splits prompt injection datasets from Hugging Face for guardrail model training.
nvtruongopsfpt-capstone-rubrics-and-process
Tracks FPT University capstone milestones, grading rubrics, and weekly process reports.
nvtruongopsguardrail-api-and-dashboard
Implements FastAPI guardrail middleware and Streamlit dashboard for prompt injection detection.
nvtruongopsllm-security-research
Guides prompt injection and jailbreak research with curated academic references and citation standards.
nvtruongopsacademic-terminology-glossary
Detects academic terms in technical reports and generates a four-column glossary with anchored links.
nvtruongopsguardrail-evaluation-metrics
Computes classification metrics, adversarial robustness scores, and latency benchmarks for prompt injection guardrails.
nvtruongopsresource-and-literature-validation
Validates URLs, YouTube videos, and academic citations while retrieving open-access PDFs via DOI lookup.
Frequently Asked Questions About Truong Nguyen
FAQPage SchemaWhat tasks can I accomplish with nvtruongops's PI-Guard skills?▼
You can train TF-IDF, XGBoost, and Transformer classifiers for prompt-injection detection, engineer balanced jailbreak datasets, compute F1/FPR/ROC-AUC metrics, build a FastAPI guardrail proxy with a Streamlit dashboard, and produce the FPT University IAP491 capstone thesis with IEEE citations.
Who are these PI-Guard skills designed for?▼
They target the PI-Guard capstone team at FPT University: ML engineers fine-tuning DeBERTa-v3 guardrail models, security researchers studying OWASP LLM01 and NIST AI 100-2e2025 threat models, and students preparing Review 1 defenses and graduation thesis chapters.
How does the team Git synchronization and audit process work?▼
Members work only inside personal workspaces/ directories while the leader (nvtruongops) holds exclusive merge rights to shared folders. Automated audits inspect commit history, git diffs, staging areas, and pull requests to detect and block permission violations or changes to immutable files.
What guardrail baselines and research sources are covered?▼
The skills survey SOTA guardrails including ProtectAI DeBERTa, Lakera Gandalf, Llama Guard, and NeMo Guardrails, with literature strictly from 2022 onward. References are validated via OpenAlex and Semantic Scholar, fetching open-access PDFs by DOI to guarantee zero dead links and no paywalled sources.
How is the documentation portal deployed and maintained?▼
Thesis, reports, and references are synchronized from the repository to a MkDocs Material web portal, with automated link and citation-anchor integrity checks, then deployed to the GitHub gh-pages branch for publishing.