Lambenthan/empiricalwiki / i18n/en/skills/discover
discover
VerifiedBuild 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.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.3 KB |
| references/ranking-signals.md | markdown | 5.8 KB |
| references/seed-modes.md | markdown | 2.5 KB |
| references/wiki-dedup.md | markdown | 2.4 KB |
references/wiki-dedup.md
View on GitHub# /discover wiki dedup
`tools/discover.py` deduplicates candidates against the existing wiki when `--wiki-root wiki` is passed. This document explains what the dedup does and does not catch, so the user-facing report is accurate.
## What it catches
For each candidate, `tools/discover.py` extracts the `arxiv_id` from the candidate record (S2's `externalIds.ArXiv`, DeepXiv's `arxiv_id`, etc.) and checks whether any existing `wiki/papers/*.md` page has a matching `arxiv` (or legacy `arxiv_id`) in its frontmatter. Matches are dropped from the shortlist before scoring; the count is reported as `wiki_dedup_count`.
This catches the typical case: an already-ingested paper bubbles up as a recommendation again. Surfacing such a paper would waste the user's review attention; dropping it is correct.
## What it does not catch
- **Title-only matches**: a paper in the wiki without `arxiv` or `arxiv_id` (e.g., a journal article ingested via `/edit`) will not match a candidate by title alone. This is intentional — fuzzy title matching produces false positives that hide legitimate candidates.
- **arXiv version skew**: `2106.09685` and `2106.09685v3` should both be treated as the same paper. The frontmatter scanner strips `arxiv:`/`ARXIV:` prefixes but does not currently strip `vN` suffixes. If you find duplicates leaking through, normalise the version suffix in the candidate's `arxiv_id` before comparison.
- **Cross-source duplicates within the candidate set**: the dedup pass before wiki filtering uses `_candidate_key` (arxiv → S2 paperId → title-slug) which catches most cross-source duplicates from S2 and DeepXiv. Fully missing IDs and titles are dropped silently.
## What to do with a "high dedup" report
If `wiki_dedup_count` is high relative to `candidates_total` (e.g., 30 / 50), the wiki is already well-covered for these anchors. Two interpretations:
1. The user is looking for breadth and should switch to a different seed (different anchor, broader topic, or `--from-wiki` to explore adjacent papers).
2. The recommendation channel is genuinely saturated — there is little new to recommend in this neighborhood.
The skill should mention high dedup in the user-facing report; do not hide it.
## What dedup does not do
`/discover` never modifies the wiki to "fix" a duplicate. If the candidate's metadata seems richer than what is currently in the wiki, that is a `/edit` or `/check` concern, not a `/discover` concern.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 472k | Updated | Claude CodeCodex | Verified | Install |
| 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: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| 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 available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |