Skill bundles

claude mcp add --transport http particle-podcasts "https://mcp.particle.pro?tools=particle_entity_resolve,particle_podcast_resolve,particle_podcast_search_transcripts,particle_podcast_list_episodes,particle_podcast_get_episode"

---
name: particle-podcast-intelligence
description: Search and analyze podcast dialogue via the Particle Pro MCP server — resolve shows and entities to slugs, search transcripts semantically, list episodes by filters, and pull full episode detail.
---

# Particle Podcast Intelligence

Tools: particle_entity_resolve, particle_podcast_resolve, particle_podcast_search_transcripts, particle_podcast_list_episodes, particle_podcast_get_episode.

## Workflow

1. **Resolve names to slugs first — never guess a slug.** Free-text name of any
   kind → `particle_entity_resolve` (returns a typed handle: person, company,
   place, other). Show title, iTunes ID, or RSS URL → `particle_podcast_resolve`.
2. **Search by intent.** Dialogue *about* a topic → `particle_podcast_search_transcripts`
   (`semantic_search` for paraphrase, `keyword_search` for exact phrases, both for
   hybrid ranking; at least one is required). Filter-driven discovery (by podcast,
   person, company, language, date, duration) → `particle_podcast_list_episodes`.
3. **Drill in.** `particle_podcast_get_episode` with `include: ["segments", "entities",
   "clips", "topics", "transcript"]` — request only the sections you need; the
   default response is lean.

## Conventions

- Every slug a tool returns is a valid input to the other tools — resolve once,
  then traverse.
- `particle_entity_resolve` returns a *typed* handle: feed a `person` result's
  slug to `person_slug` parameters, a `company` result's slug to
  `company_slug`, and a `place`/`other` result's slug to `entity_slug` (the
  catch-all filter on search and list_episodes). Don't cross the streams — a
  place slug in `person_slug` resolves nothing.
- Search matches return bounded transcript windows; raise `context` to widen them
  in place instead of fetching the full transcript.
- There is no separate clip-search tool: relevant clips arrive inline on search
  matches, and a known episode's full clip list is `get_episode` with
  `include: ["clips"]`.
- `particle_podcast_search_transcripts` is a premium-tier tool; the others here
  are standard.

claude mcp add --transport http particle-companies "https://mcp.particle.pro?tools=particle_company_resolve,particle_company_get,particle_person_get,particle_podcast_find_mentions,particle_company_get_podcast_ad_presence"

---
name: particle-company-deep-dive
description: Research a company via the Particle Pro MCP server — resolve by name/ticker/domain, pull the bundled profile with people/products/competitors, trace podcast mentions, and map its podcast sponsorships.
---

# Particle Company Deep Dive

Tools: particle_company_resolve, particle_company_get, particle_person_get, particle_podcast_find_mentions, particle_company_get_podcast_ad_presence.

## Workflow

1. **Resolve.** `particle_company_resolve` takes a free-text name (`query`, one
   company at a time) or a `ticker`/`domain`/`cik`/`qid` — those four accept
   comma-separated values for bulk lookup — and returns the canonical
   `company_slug`.
2. **Profile.** `particle_company_get` with `include: ["people", "products",
   "competitors"]` — each person in the response carries a `person_slug`.
3. **Follow the people.** `particle_person_get` with
   `include: ["podcast_appearances", "companies"]` turns a slug from step 2 into
   role history and recent episodes (with episode/podcast slugs for follow-up).
4. **Trace the conversation.** `particle_podcast_find_mentions` with
   `company_slug` finds dialogue lines naming the company. Default
   `format: "summary"` previews ~10 mention lines per episode for scanning;
   call back with `format: "detail"` plus `episode_slug` (up to 10,
   comma-separated) for complete mention windows with context.
5. **Map the ad footprint.** `particle_company_get_podcast_ad_presence` shows
   where the company appears as a sponsor.

## Conventions

- Slugs are edges: company → its people → a person's appearances → an episode's
  mentions, all in four calls.
- Mentions and ad-presence are premium-tier tools; resolve/get are standard.
- A 404 on a slug means it's stale or misspelled — re-resolve it, don't retry.

claude mcp add --transport http particle-people "https://mcp.particle.pro?tools=particle_person_resolve,particle_person_get,particle_topic_browse,particle_podcast_list_guests,particle_podcast_get_guest,particle_podcast_find_mentions"

---
name: particle-guest-research
description: Research podcast guests and people via the Particle Pro MCP server — resolve people to canonical slugs, pull profiles and appearance histories, browse the guest directory and trends, and trace mentions.
---

# Particle Guest & People Research

Tools: particle_person_resolve, particle_person_get, particle_topic_browse, particle_podcast_list_guests, particle_podcast_get_guest, particle_podcast_find_mentions.

## Workflow

1. **Resolve.** `particle_person_resolve` turns a free-text name (comma-separated
   for bulk) into the canonical `person_slug` every other tool accepts.
2. **Profile.** `particle_person_get` with `include: ["external_links",
   "podcast_appearances", "companies"]` for bio, links, recent appearances, and
   role history.
3. **Guest analytics.** `particle_podcast_get_guest` (same slug as `guest_slug`)
   for appearance stats and per-podcast rollups. `particle_podcast_list_guests`
   with `mode: "directory"` ranks by lifetime appearances; `mode: "trends"`
   surfaces who's appearing across multiple shows right now; `topic_slug`
   (a slug from `particle_topic_browse`) narrows either to a topic.
4. **What's said about them.** `particle_podcast_find_mentions` with
   `person_slug` — `format: "summary"` to scan (~10 lines per episode),
   `format: "detail"` + `episode_slug` for the full mention windows.

## Conventions

- `person_slug` and `guest_slug` are the same identifier — resolve once, use it
  everywhere.
- Some people lack a knowledge-graph link; mention lookups for them return an
  explained empty result, not an error — report that, don't retry.
- `find_mentions` is premium-tier; the rest here are standard.