Moonlight Lupin Agent Skills - Setup Guide¶

Source: moonlight-lupin/agent-skills (88⭐, 15 forks, MIT LICENSE; actively maintained) Skill family: moonlight-lupin/agent-skills (36 SKILL.md files in-repo - 35 installable skills + 1 test fixture; 37 indexed listings) Combined Installs: ~3,354 across indexed listings (Oct 4, 2026 snapshot) Category: Hermes Agent Ecosystem / Skill Suite Quality Tier: 🟡 Beta (CI-tested, versioned skills from a single maintainer; community validation still forming - verified Oct 4, 2026)

A collection of 35 AI agent skills built specifically for Hermes Agent, organised by domain: research, agent-ops, productivity, creative, mlops, devops, and web-scraping, plus two installable plugins (skill-retrieval and lumen). The suite covers the full agent lifecycle: onboarding a customer onto Hermes (hermes-onboarding), auditing token spend (input-token-analysis, input-token-overheads), running source-grounded research (deep-research, fact-checker, source-tracker), and packaging the output (marp-deck, clips-studio). Parked on the Sep 29, 2026 sweep as sub-500 and un-parked Oct 4, 2026 after a growth re-check.

Installation¶

# Full suite (all 35 skills)
npx skills add moonlight-lupin/agent-skills

# Or install a single skill
npx skills add moonlight-lupin/agent-skills --skill deep-research

The repo also ships two installable Hermes plugins under plugins/; the README documents installing them from the plugin catalog or by copying/symlinking into your profile's plugins/ directory:

hermes plugins install skill-retrieval

Roster - All 35 Skills¶

Skill	Domain	Installs	What It Does
deep-research	research	119	Iterative research engine: structured evidence, source-quality ranking, refute queries, deterministic validation gates
fact-checker	research	114	Claim verification with confidence ratings and cited reports
website-scraping	web-scraping	112	Recon to lightest-tool extraction to JSONL with run manifest
disk-cleanup	devops	110	Disk triage: survey mounts, safe/ask buckets, execute approved set, verify delta
log-analyzer	agent-ops	107	Log pattern detection: error clusters, rate limits, timeouts, tool failures
file-organizer	productivity	107	LLM-powered directory organizer: scan, propose, confirm, chunked moves
source-tracker	research	106	Persistent citation database: dedup, topic tags, link health, bibliography export
skill-maintainer	agent-ops	106	Skill library maintenance: author, curate, upstream drift tracking, publish
pexels-stock-photos	creative	106	Free real-world stock photos via Pexels API with attribution handling
news-monitoring	research	104	Recurring news digests with cron delivery, dedup, and multi-language output
youtube-topic-research	research	104	YouTube search, transcript, and summary pipeline
decision-log	productivity	103	ADR-style decision journal with superseding chains and cron reviews
task-brief	productivity	103	Goal/context/constraints brief compiled and confirmed before substantial tasks
model-compare	mlops	103	Blind multi-model A/B comparison plus embedding benchmark
entity-research	research	103	Cited company/person dossiers: ownership, adverse media, sanctions, litigation
endpoint-probe	research	103	API/MCP surface discovery across REST, GraphQL, SOAP, and JSON-RPC
claude-plugin-converter	agent-ops	103	Convert Claude Code plugins into self-contained Hermes plugins
scheduled-summary	productivity	102	Cron-driven cross-session digests for messaging platforms
media-analyzer	research	102	Rhetorical technique detection: loaded language, framing, omission
travel-itinerary	productivity	101	Business-trip itineraries from emails/PDFs to Markdown + .ics + chat variants
people-enrichment	research	101	Person/company enrichment via PDL, exported to styled .xlsx
notebooklm-mode	research	101	Source-grounded Q&A from a source vault (strict or augmented grounding)
hermes-onboarding	agent-ops	101	21-step customer onboarding: gateway, dashboard, memory, search, guardrails, crons
clips-studio	creative	101	Staged fal.ai short-video generation: text-to-video, animate still, camera moves
fill-template	productivity	100	Bulk-fill Word/Excel templates from a data table (mail-merge)
library-rag	research	99	Semantic search over a personal library (Nemotron-3-Embed + sqlite-vec)
image-studio	creative	99	Staged fal.ai image generation/editing with cost logging and --dry-run
input-token-overheads	agent-ops	98	Audits per-turn input token cost: measure, rank, reduce
input-token-analysis	agent-ops	74	Explains and fixes input-token spend (pairs with input-token-overheads)
operator-brain	agent-ops	50	8-module behavioral stack: act, lead with outcome, ground every claim
receipt-compiler	productivity	49	Phone-camera receipts to straightened B&W scans to expense-claim PDF
marp-deck	productivity	6	Marp presentations: Markdown to PDF/PPTX/HTML with themes and render tests
skill-retrieval	plugins	139	BM25 skill-retrieval plugin: injects only the top-K relevant skills per turn
wiki	plugins	9	Lumen wiki plugin: OKF frontmatter, three-layer pages, ingest/query/lint workflows
curator	plugins	9	Lumen wiki plugin: wiki_curator script + memory-source adapters

Plugins¶

Two installable Hermes plugins ship in-repo. skill-retrieval replaces the full per-turn skill block with BM25 retrieval (top-K injection, stdlib only, per-profile index caching, optional rerank). lumen turns conversation history and memory into a maintained markdown wiki (sessions-first, cap-bounded curation, source-attribution contract). Both include tests runnable with python3 -m pytest.

CorpusIQ Use Cases¶

Use Case	How
Hermes fleet reference	hermes-onboarding (21-step customer onboarding), input-token-overheads, and skill-retrieval map directly onto the operational stack we run for Hermes deployments - use as a second opinion on our own setup decisions
Research discipline cross-check	deep-research, fact-checker, and source-tracker implement the same source-grounded discipline we apply in intelligence work: evidence-basis labels, refute queries, validation gates
Plugin packaging pattern	lumen and skill-retrieval are clean examples of the Hermes plugin layout (plugin.yaml + skills + stdlib tests) for our own plugin development

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations / Verification¶

Verified Oct 4, 2026: 36 SKILL.md files via the GitHub trees API (branch main); 88⭐ / 15 forks; MIT LICENSE in-repo; repo created Jul 8, 2026 and actively maintained.
Single-maintainer provenance (personal project). Despite CI, per-skill tests, and versioned skills, community validation is still forming - treat as supervised-use until you have reviewed the skills you enable.
skills.sh also indexes one test fixture (greet, 48) and one listing with no current in-repo SKILL.md (document-converter, 52); both are excluded from the roster but included in the combined-install figure.
Some skills depend on external services: fal.ai keys for image-studio and clips-studio, Pexels and PDL API keys, tesseract for receipt-compiler; plugin skills assume a Hermes runtime with plugins enabled.
No live install test was performed; install counts are the Oct 4, 2026 skills.sh sweep snapshot.

Security¶

skills.sh per-skill verdicts (sampled 3 skills, verified Oct 4, 2026) - verdicts vary per skill; check the skill's security page on skills.sh before production use:

Skill	Gen Agent Trust Hub	Socket	Snyk
deep-research	Pass	Pass	Warn
skill-retrieval	Pass	Pass	Warn
hermes-onboarding	Pass	Warn	Pass

← Skills Catalog | Skills Marketplace → Powered by CorpusIQ