Haiyang Jiang
Community@ocean326
I 'm Haiyang Jiang from BeiHang University, loving coding.
Agent Skills by Haiyang Jiang
Showing 66 vetted skills indexed across 1 GitHub repositories.
transferrec-literature-chain
Query local trajectory-representation literature and write new materials into the progressive-access library.
transferrec-automation-governor
Enforce single bounded actions per round in Transfer_Recovery automation loops.
thinking-lenses
Route prompts through reasoning lenses and generate compact execution-ready prompts.
proposal-synthesis
Frame ambiguous goals into 2-5 proposal briefs with a structured proposal_packet.
execution-router
Route tasks to local, remote SSH, or subagent execution lanes.
huangchao
Evaluate agent workflows using a HKUDS-style product and interface lens.
delivery-router
Route ambiguous delivery and product tasks to appropriate sub-skills.
searxng-search
Search the web locally via a SearXNG stack without API keys.
local-docker-packaging
Automate local Docker packaging by building images and exporting tarballs.
research-training-and-ablation-loop
Analyze AI training runs and ablation results to recommend next experiments.
skill-skill-usage-close-loop
Route thread skill usage to PersonalBrain's close-loop workflow and append review evidence to the governance ledger.
test-cases
Convert product requirements into structured test cases with markdown output.
parallel-deep-research
Orchestrate parallel-cli deep-research tasks with multi-turn context and progress tracking.
research-paper-reading-compass
Convert research papers into structured notes with a fixed schema.
reddit-fetch
Fetch Reddit content via Gemini browsing with a JSON API fallback.
result
Poll a research task by run ID and retrieve its result in JSON.
parallel-data-enrichment
Enrich CSV or inline entity lists with web-sourced fields via parallel-cli.
attach-personalbrain
Attach a task to a PersonalBrain repository using AGENTS.md or catalog.jsonl anchors.
architect-router
Route architecture inquiries to the appropriate architecture lane.
parallel-web-extract
Extract text and data from URLs, PDFs, and JavaScript-heavy sites.
source-to-skill
Distill mixed source material into a skill candidate packet with task_signature, taste_rubric, and operator_rules.
codex-session-evolution
Analyze historical Codex session logs to identify recurring workflow and prompt patterns.
research-experiment-design-planner
Design compute-aware AI experiment plans with baselines, metrics, and ablation matrices.
dispatching-parallel-agents
Dispatch independent tasks to specialized agents with isolated contexts.