Instructions file CodexOpenCode
AGENTS.md instructions for Greyman-Seu/FollowHub, covering repository guidelines, project structure & module organization, build, test, and development commands, coding style & naming conventions and testing guidelines.
Skill Claude CodeCodex
Use when an agent needs raw arXiv daily or backfill data before any semantic filtering, especially category-wide New submissions for the FollowHub arXiv pipeline.
Skill Claude CodeCodex
Use when the user asks for today's arXiv follow-up or missed-day arXiv backfill; this skill is an agent-native workflow that must execute collect -> title-prefilter -> filter -> enrich -> publish in order.
Skill Claude CodeCodex
Use when arXiv result files need stable summary, affiliation, score, and link fields, or when arxiv-find needs its default post-retrieval enrichment phase.
Skill Claude CodeCodex
Use when the user wants to find architecture figures, system overview diagrams, or model network structure images from an arXiv paper. Trigger on requests like "find architecture diagram", "extract model structure figure", "get system overview image from paper", or when given an arXiv ID with figure-related keywords.
Skill Claude CodeCodex
Use when an agent or subagent needs to semantically filter arXiv papers for FollowHub, deciding inclusion, Follow domains, one-line Chinese summaries, Chinese abstracts, and review reasons.
Skill Claude CodeCodex
Use when the user wants to search arXiv papers by keywords or topic context, generate daily briefs, or backfill missed daily briefs while keeping each day separate and limited to New submissions.
Skill Claude CodeCodex
Use when an agent needs a fast title/category-only screening pass before full arXiv filtering, deciding keep/drop/uncertain for batches of raw papers.
Skill Claude CodeCodex
Use when one or more arXiv papers should be turned into high-quality llm-wiki entries through the arxiv-to-wiki, paper-analyze, and llm-wiki pipeline, with post-write completeness checks and default publication to the R2 package and public wiki page unless the user explicitly opts out.
Skill Claude CodeCodex
Use when the user wants a static HTML viewer for arXiv daily briefs, backfills, or search results produced by arxiv-find, including local favorites and clipboard export of selected arXiv IDs.
Skill Claude CodeCodex
Use when the user asks to run, publish, backfill, or check today's FollowHub daily. Run arXiv and RSS on weekdays, RSS only on Saturday and Sunday, and publish the completed results to the Follow page unless the user explicitly limits the request to one source.
Skill Claude CodeCodex
Use when FollowHub needs to turn daily follow digests or arXiv result files into page-ready Follow data artifacts, sync them into the page submodule, or prepare publishable JSON bundles for Follow pages.
Skill Claude CodeCodex
Use when the user wants a local Markdown file rendered into temporary HTML and served on a local preview URL for quick review before publishing or syncing.
Skill Claude CodeCodex
Use when the user wants one paper deeply analyzed into a final Markdown note, with stable sections, optional figure references, and direct output into an llm-wiki source directory configured in YAML.
Skill Claude CodeCodex
Use when one analyzed wiki source note should be exported and published for immediate remote reading, including rendered HTML and source JSON upload to R2.
Skill Claude CodeCodex
Use when FollowHub needs a single workflow to move from analyzed wiki source notes to refreshed website data and publish-ready wiki pages, including sync and validation steps.
Skill Claude CodeCodex
Use when uploading, syncing, listing, or deleting files in FollowHub's Cloudflare R2 bucket, especially when credentials live in /.followhub/config.yaml or FOLLOWHUBCONFIG and the agent should use the bundled R2 rclone helper.
Skill Claude CodeCodex
Use when deduped RSS items need lightweight story grouping across same-day and cross-day repeated coverage, while preserving agent handoff for ambiguous event merges.
Skill Claude CodeCodex
Use when an agent needs raw RSS entries for daily collection before semantic filtering, especially for WeChat, X/Twitter, and other feed-based sources.
Skill Claude CodeCodex
Use when the user asks for today's RSS follow-up or a missed-day RSS backfill; this skill is an agent-native workflow that must execute collect -> normalize -> fetch -> dedupe -> cluster -> prefilter -> filter -> enrich -> digest -> publish -> verify in order.
Skill Claude CodeCodex
Use when normalized or fetched RSS items need same-content deduplication before semantic filtering, especially across mirrored WeChat/X feeds and repeated source entries.
Skill Claude CodeCodex
Use when filtered and enriched RSS items need to be grouped into a daily digest JSON bundle for FollowHub.
Skill Claude CodeCodex
Use when RSS result files need stable author, link, source, and summary fields, or when missing Chinese text should be handed off to agent completion tasks.
Skill Claude CodeCodex
Use when normalized RSS entries need URL-based content fetching before semantic filtering, especially for WeChat articles and X/Twitter thread pages.