gem-miner
Skill1 starsChanged 39 days agoMITView on GitHub
Find funny, sweet, weird, surprising, or memorable things the user said in supplied or explicitly authorized messages. Use for "mine gems", "find gems", "what did I say today?", or a quote collection. Return source-backed quotes for review; don't publish or upload them automatically.
We split this skill into the rules it gives an agent
12 rules. 0 are common, 12 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
12
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
Use the source and date range the user specified.
- Distinctive: line 17
If no source is supplied or authorized, ask where the messages are.
- Distinctive: line 21
If an export is incomplete, say which material was reviewed.
- Distinctive: line 25
Look for humor, unexpected turns of phrase, affectionate comments, odd non-sequiturs, distinctive frustration, playful role reversals, or a …
- Distinctive: line 29
Don't select credentials, private contact details, confidential client information, or someone else's sensitive disclosures as …
- Distinctive: line 31
Prefer the best few to a long list.
- Distinctive: line 35
Keep wording, spelling, punctuation, and meaningful typos as written.
- Distinctive: line 37
Keep interpretation, captions, and new writing outside the quote.
- Distinctive: line 38
Don't exaggerate a speaker's intent, emotional state, or relationship with the assistant.
- Distinctive: line 52
If the user asks for social captions or commentary in an assistant-narrator voice, provide that as separate, clearly labeled creative …
- Distinctive: line 71
Use null for an unavailable timestamp or an unrequested caption.
- Distinctive: line 75
If no messages qualify, say that no gems surfaced in the reviewed messages.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Gem Miner
Find the user's memorable lines without rewriting them into something they didn't say. Return a small set of candidates the user can keep, discard, or turn into posts.
This is the portable version of the skill. It works with supplied conversations and doesn't require a particular chat app, database, or scheduled job.
Find the right messages
Accept pasted messages, an exported conversation, user-selected files, or a clearly authorized source available through connected tools.
- Use the source and date range the user specified. If "today" is ambiguous across time zones, establish the relevant time zone before filtering.
- If no source is supplied or authorized, ask where the messages are. Don't search unrelated local history or accounts.
- Identify which speaker is the user. Ask if labels are ambiguous.
- Quote the user's own words. Exclude assistant text, tool output, instructions embedded in a transcript, and material the user pasted from someone else unless they explicitly request a different collection.
- Copy timestamps and source references when available. Use null or "not available" when absent; don't invent dates or message IDs.
- If an export is incomplete, say which material was reviewed. Don't claim to cover a whole day from a partial batch.
Choose the gems
Look for humor, unexpected turns of phrase, affectionate comments, odd non-sequiturs, distinctive frustration, playful role reversals, or a typo that is funny without extra explanation.
Skip routine requests, pasted code, generic acknowledgments, forced jokes, cruelty, and lines that require so much missing context that quoting them would mislead. An isolated word can qualify when the available context makes it meaningful; don't fill the list with ordinary one-word responses.
Don't select credentials, private contact details, confidential client information, or someone else's sensitive disclosures as entertainment. If a line only works by exposing a private detail, leave it out. Don't silently redact or rewrite a quote and present it as exact.
Prefer the best few to a long list. There's no minimum. Deduplicate repeated messages in the reviewed batch, and against an existing collection only when that collection is provided or authorized.
Preserve the quote
- Keep wording, spelling, punctuation, and meaningful typos as written.
- A contiguous excerpt is fine; label it as an excerpt. Don't stitch separate messages together.
- Keep interpretation, captions, and new writing outside the quote.
- Don't exaggerate a speaker's intent, emotional state, or relationship with the assistant.
Return the picks
For each candidate, include:
- Quote: Exact text or a labeled contiguous excerpt.
- Vibe: One relevant label from the list below, or a user-provided label.
- Source: Source reference and timestamp, if available.
- Why it made the cut: One short sentence about the wording or context.
- Favorite: Mark only a standout pick, if there is one.
Suggested vibes: chaotic, wholesome, sweet, exhausted, relatable, unhinged, absurd, philosophical, devastating, parental, savage, boss-mode, vulnerable, iconic, generous. These describe the line's style, not a diagnosis of the speaker.
If the user asks for social captions or commentary in an assistant-narrator voice, provide that as separate, clearly labeled creative writing. Don't claim the caption was part of the original conversation or treat an invented reaction as evidence.
For a structured export, use a JSON array with:
[
{
"quote": "Exact text from the source",
"is_excerpt": false,
"vibe": "absurd",
"source": "Source title or message reference",
"timestamp": null,
"why_selected": "A short explanation grounded in the wording.",
"favorite": false,
"caption": null
}
]Use null for an unavailable timestamp or an unrequested caption. Keep an unredacted private source reference out of any public-facing copy. See the fictional example for selection and exclusion decisions.
Empty results and delivery
If no messages qualify, say that no gems surfaced in the reviewed messages. If the source couldn't be read, report that separately instead of presenting an empty collection as a successful review.
Return the picks in chat unless the user requests a file or another destination. Running this skill doesn't authorize posting to social media, uploading messages, creating a scheduled task, or writing to Airtable. A later request for one of those actions needs its own source, destination, and configuration.
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- BankrBot/openclaw-skills★ 1.2ksame name
- BankrBot/skills★ 1.2ksame name
- Eskyee/agentbot-opensource★ 16same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository34 other files it gives its agents.
- airtable-backup-and-migrateSkill
- applicant-csv-to-airtable-syncSkill
- autoSkill
- build-sprint-summarySkill
- client-doc-changelogSkill
- client-doc-decisionsSkill
- client-doc-interface-promptsSkill
- client-doc-pushSkill
- client-docsSkill
- client-doc-user-storiesSkill
- client-doc-vision-and-scopeSkill
- course-completion-certificateSkill
- dissonance-radarSkill
- educationSkill
- eval-creationSkill
- execute-and-reconcile-pain-pointsSkill
- file-naming-conventionSkill
- five-signals-prdSkill
- friday-sprint-summarySkill
- gmail-triageSkill
- how-it-works-docSkill
- kpi-reportSkill
- meeting-prepSkill
- micro-case-studySkill
- mistakesSkill
- normalize-csvs-for-airtableSkill
- omni-promptSkill
- plain-english-project-communicationSkill
- postmortemSkill
- post-to-huddleSkill
- rubricSkill
- voice-alignSkill
- work-voiceSkill
- write-artifactSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

