Skip to content

Lambenthan/empiricalwiki / i18n/en/skills/discover

discover

Verified

Build a ranked shortlist of candidate papers (anchor-driven, topic-driven, or derived from current wiki state) that the user — or an upstream skill — may decide to feed into `/ingest`. Use whenever the user asks "what should I read next", "find papers similar to this one", "recommend related work", "what's around this topic", or whenever `/ingest` is invoked with `--discover`. Does not ingest; only proposes.

FileKindSize
SKILL.mdskill md10.3 KB
references/ranking-signals.mdmarkdown5.8 KB
references/seed-modes.mdmarkdown2.5 KB
references/wiki-dedup.mdmarkdown2.4 KB

references/seed-modes.md

View on GitHub
# /discover seed modes

Pick exactly one mode per invocation. The decision is based on what the user (or calling skill) actually said, not on what the wiki contains.

## Anchor mode (`from-anchors`)

Use when the user named one or more specific papers, or when this is a post-`/ingest` `--discover` follow-up.

Triggers:

- "find papers similar to LoRA"
- "what's related to this one I just ingested"
- one or more arXiv URLs / IDs / wiki paper slugs in the request
- `/ingest --discover` invocation (anchor = the just-ingested paper's arXiv ID)

Anchor mode is the strongest signal channel — Semantic Scholar's recommendations endpoint returns semantically similar papers based on its trained model, which is more useful than keyword search when the user has a concrete reference point.

If the user supplies negatives ("not these", "different from X"), pass them via `--negative`. The S2 recommendations endpoint pushes the result distribution away from negative anchors, which is useful when the user wants to escape a sub-area they already know.

## Topic mode (`from-topic`)

Use when the user gave a topic, direction, or set of keywords without naming specific papers.

Triggers:

- "find papers about diffusion model fine-tuning"
- "what's been written on retrieval augmented generation"
- a domain phrase with no anchors

Topic mode runs S2 search and (when available) DeepXiv search, then ranks. It is a lighter alternative to `/init`'s planner: useful for exploration but **not** a replacement for `/init`'s broader bootstrap workflow. If the user wants to seed a fresh wiki with a topic, route them to `/init` instead of bulking up `/discover`.

## Wiki mode (`from-wiki`)

Use when the user asked open-ended "what should I read next" with no anchor and no topic.

Triggers:

- "give me the next batch of papers to read"
- "what's a good follow-up to my current wiki"
- explicit `--from-wiki` flag

Wiki mode picks the wiki's most recently modified paper pages, extracts their arXiv IDs, and uses them as anchors. This implicitly biases discovery toward whatever the user has been working on lately — usually the desired behavior.

If `wiki/papers/` is empty or no papers carry an `arxiv` or `arxiv_id` frontmatter field, wiki mode cannot run. Tell the user the wiki is too sparse and suggest topic mode (or `/init`).

## What if the user gave both an anchor and a topic?

Prefer anchor mode. Anchors are a much stronger signal than a topic string. Mention the topic in the user-facing report so they know it was noted, but the discovery itself runs through `from-anchors`.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall