reddit-search git:20260817.d6bd9d1A

Immutable. This exact content never changes and is served at /api/v1/blob/185a87f29586d4cd.

---
name: reddit-search
description: Research public Reddit posts, comments, communities, post threads, and user profiles with bounded collection. Use for audience voice, pain points, product opinions, subreddit discovery or feeds, comment deep-dives, and explicitly requested public profile activity.
metadata:
  postplus:
    familyId: reddit
    familyName: Reddit
---

# Reddit Search

Use this skill for bounded public Reddit research. Apply the shared rulebook
and user-guidance rules from `postplus-shared`.
When a supported command completes but evidence is empty, sparse, noisy,
off-topic, or the wrong record type, apply the `postplus-shared` reference
`research-quality-recovery.md`; hard execution errors still fail fast.

## Experience Rules

1. Speak in the user's business language, not collection terminology.
2. Ask at most one question, and only when it changes the route or cost bound.
3. Start with the smallest useful sample and expand only after review.
4. Keep implementation, delivery, and network controls out of the conversation.
5. Fail fast on hard command, contract, auth, or private-surface failures. When
   a supported run completes with empty, sparse, noisy, or off-topic evidence,
   apply the shared bounded research-quality recovery rule before concluding
   that the evidence is exhausted.

## Route Index

Read `references/shared-contract.md` for every route, then the route reference
and `references/result-shapes.md`. All references are direct from this file.

| User asks for | Route references | Default first pass |
| --- | --- | --- |
| Keyword posts, communities, or a Reddit search URL | `references/search.md` | 20 posts or 10 communities |
| Keyword comment evidence | `references/search.md`, then `references/post-comments.md` | 20 discovery posts, then up to 3 relevant threads |
| One supplied post URL's comments | `references/post-comments.md` | 1 post and up to 50 comments |
| A guided pain-point or product-opinion deep-dive | `references/search.md`, then `references/post-comments.md` | 20 discovery posts, then 3 suggested threads |
| A subreddit search, feed, metadata, or deep collection | `references/community.md` | 20 feed posts; deep scope is bounded separately |
| An explicitly named public Reddit profile and its activity | `references/profile.md` | 20 posts and 20 comments |
| Normalizing, deduplicating, or presenting any result | `references/result-shapes.md` | Preserve the raw result file |

## First Question

Do not ask when the user supplied a keyword, URL, subreddit, or profile and a
safe default applies. Otherwise ask only the highest-information question:

| Missing decision | Ask |
| --- | --- |
| No research target | `What Reddit topic, URL, subreddit, or public profile should I research?` |
| A deep subreddit request has no bound | `What date range or maximum post count should bound the deep collection?` |
| Multiple expensive directions compete | `Which one should I run first: discovery, a thread deep-dive, a subreddit, or a public profile?` |

Never ask for credentials, collection keys, schema fields, implementation
choice, delivery settings, or retry strategy.

## Run Protocol

1. Select one route and apply its references.
2. Write the raw request object to a local file under `.postplus/`.
3. Run the narrowest collection that answers the first pass.
4. Inspect record types and relevance. If the evidence is not useful, apply the
   recovery flow in `references/shared-contract.md` without repeating an
   identical request or exceeding the approved cost bound.
5. Keep the complete raw records in the result file; normalize only the
   user-facing evidence according to `references/result-shapes.md`.
6. Report scope, counts by result type, representative evidence, limits, and
   one useful next action. Do not imply a bounded pass is exhaustive.

Multiple keywords are separate, attributable requests. They may run in
parallel, but their result sets remain separate until presentation.

## Public Command Boundary

- Readiness: `postplus doctor --skill reddit-search`.
- Inspect an unknown request contract only when needed with
  `postplus research schema --collection-key <collection-key> --json`.
- Select the route-owned collection key:
  - `reddit-search` for keyword discovery, public search URLs, community search,
    and ordinary subreddit feeds;
  - `reddit-post-comments` for one or more selected post threads;
  - `reddit-subreddit-posts` for bounded deep subreddit collection;
  - `reddit-user-activity` for an explicitly named public profile.
- Collect with
  `postplus research collect <collection-key> --request <input.json> --output <result.json>`.
- The request file is the raw collection input object, never a hosted envelope.
- If a run returns a saved async checkpoint, resume it with
  `postplus research collect --resume-from <result.json>` rather than starting
  a duplicate run.
- If a quote-confirmation challenge appears, show its scope and price. Only
  after the user confirms, run
  `postplus quote confirm --json --challenge-file <challenge.json>` and retry
  with the returned token.
- If an owned CLI command fails, report the exact error and stop. Do not use
  unpublished tools, payload rewrites, or another service as a fallback.

## Scope Boundary

Supported: public keyword discovery, public search URLs, public post threads,
public subreddit content and metadata, and explicitly requested public profile
activity.

Excluded: login or account access, posting or messaging, private/deleted
content recovery, real-identity inference, private-data enrichment, automated
sentiment analysis, media download, external delivery, and network tuning.

<!-- BEGIN GENERATED EXECUTION EXAMPLE -->
```bash
postplus research collect reddit-search --request request.json --output result.json
```
<!-- END GENERATED EXECUTION EXAMPLE -->