Linkeddit

View saved version →

---
name: competitor-intelligence-brief
description: Turn Linkeddit competitor monitors, briefs, switching signals, and page snapshots into an evidence-backed competitive update or battlecard when users ask about competitor moves or market changes.
---

# Competitor intelligence brief

Use `list_competitor_monitors` to resolve the competitors and monitor IDs available to this account. A tracked competitor is not access to that company's private account. Use the current connected tool schemas and avoid inventing missing IDs or records.

For one competitor, retrieve `get_competitor_brief`. For a weekly overview, start with `get_weekly_brief`, using the requested week if supplied. Add `get_switching_signals` or `get_competitor_changes` only when they help answer the request. Respect the actual returned window, generation date, limits, and missing data; a latest brief may be older than the requested period.

Distinguish competitor announcements, buyer complaints, observed changes, and your interpretation. For page comparisons, compare snapshots of the same URL at different timestamps. A changed hash establishes a page changed; it does not prove a specific price or feature changed. Summary snippets may be truncated; request the tool's raw mode where available before making precise comparisons.

Deduplicate repeated reports of the same event. A single complaint or switching signal is evidence of that instance, not a market-wide trend or confirmed purchase. Report conflicting evidence and the number and independence of sources where known.

Produce the requested brief or battlecard with dated evidence links, implications for the user's product, and suggested responses. Separate verified product facts from positioning ideas. If no current brief exists, say so rather than inventing one. Do not create monitors, contact prospects, or publish comparisons as a side effect of producing a brief.

View saved version →

---
name: get-started
description: Orient a new Linkeddit user. Use when someone first connects Linkeddit or asks what it can do, how to start, or where their data is.
---

# Getting started with Linkeddit

Linkeddit is demand and competitor intelligence. It measures where AI answer
engines recommend a competitor instead of the user, tracks competitor moves,
and provides keyword research. Every result comes from data
Linkeddit recorded, so cite what the tools return rather than filling gaps.

## First steps

1. Call `answer_projects_list` to see which domains the user already tracks.
   If there are none, offer to create one with `answer_project_create`, which
   needs the user's domain, brand name and the competitors they care about.
2. For a tracked project, call `answer_metrics_get` for the headline numbers
   and `answer_gaps_list` for the buying questions where an engine recommended
   someone else.
3. For competitors, call `list_competitor_monitors`, then
   `get_competitor_brief` for the one the user names.
4. For keyword ideas, call `keyword_autocomplete` with the user's seed query.

Saved-lead search, Reddit profiles and lead insights, lead pipeline execution
and results, raw Reddit JSON fetching, and DM drafts are currently disabled.
Do not attempt those workflows or infer an account permission problem from their absence.

## Before anything that spends credits or changes something

- Always call `answer_audit_estimate` and show the user the credit estimate
  before `answer_audit_start`. Start an audit only after the user agrees.
- Publishing, unpublishing, archiving and share links change what other
  people see. Confirm with the user before calling those tools.

## Access

A connected Linkeddit account with access to the requested tools is required.
If tools are unavailable, explain that access could not be established and ask
the user to check their connection and account permissions. Do not infer a
specific subscription or authentication failure from missing tools alone.

View saved version →

---
name: keyword-opportunity-plan
description: Build a prioritized keyword and content opportunity plan from Linkeddit keyword jobs, search results, and connected search performance data when users ask what buying queries or pages to work on next.
---

# Keyword opportunity plan

Establish the product, intended buyer, and target market from the request. Retrieve existing research with `list_keyword_jobs` and `get_keyword_results`, using returned job IDs. Prefer completed jobs and identify the research date. Returned limits and filtered rows are not the complete keyword universe.

When the user wants first-party search evidence, resolve a visibility project with `answer_projects_list` and use `gsc_queries_get` if connected. Keep the requested days, query impressions, clicks, and average position together. A missing connection is not zero demand. Search performance data does not by itself prove conversions or revenue.

Use `keyword_serp` for selected queries to examine result titles, related searches, and questions. Use `keyword_autocomplete` for expansion when useful. These tools accept a query, not a locale argument in the inspected source. Do not claim results are localized or supply unsupported geography parameters. Suggestions are ideas, not verified search volume.

Cluster queries by buyer task and intent. Keep stored priority scores and conversion-potential estimates distinct from observed demand and your own judgment. Prioritize relevance to the user's product, evidence of demand, plausible page fit, and current coverage. Do not invent traffic forecasts, difficulty scores, or search volume. Cite the returned source links or job IDs.

Return a practical opportunity list: query cluster, buyer intent, evidence, proposed page type, and reason to prioritize. Treat a suggested landing page as a proposal unless its existence and relevance are verified. For a requested content brief, derive the questions to answer and evidence to include; do not fabricate competitor claims. Creating research jobs, publishing pages, and scheduling work require a separate user request and available tools.

View saved version →

---
name: prioritize-saved-leads
description: Prioritize saved Linkeddit leads against a product and buyer profile when a user asks which prospects to review, why a lead fits, or what evidence supports a relevant outreach draft.
---

# Prioritize saved leads

## Availability check

Saved-lead tools are currently disabled in the hosted connector. First check whether `search_leads` is available in the connected tool list. If it is absent, explain that this workflow is unavailable and stop; do not call it, use profile or insight tools as a workaround, invent leads, or diagnose an account permission problem. The instructions below apply only if the publisher re-enables the tools.


Use the user's product and buyer criteria. If product context is missing, `list_projects` can retrieve lead-project positioning; these are different from visibility projects returned by `answer_projects_list`. Clarify ambiguous product choices.

Use `search_leads` with supported filters: query, subreddit, status, min_score, and limit. Its query matches username, bio, audience, and industry; do not assume it searches every pain statement. Results are ranked by contactability, not purchase probability. No date filter is defined in this source; for a requested period, use returned timestamps if present and disclose when the available results cannot establish complete coverage. Never invent a supported filter.

For selected Reddit leads, `get_lead_insights` returns generated interpretations. Verify material claims using `get_user_posts` or `get_user_comments` when available and relevant. These tools take the returned username. Use public discussion evidence to identify stated business needs; avoid inferring personal or sensitive traits from a profile. `get_user_profile` is optional when its business context is actually needed.

Rank the retrieved candidates by explicit product fit, evidenced need, recency when known, and blockers. Preserve the stored contactability score separately from your reasoning. Show the source, evidence date when available, and why each candidate merits review. Say when results are a limited sample. A scraped profile or generated insight does not establish consent, a verified identity, or current buying intent.

Draft outreach only when requested, using supported statements and a useful response to the person's actual question. Do not send messages, post replies, change lead status, or claim contact occurred. If there are no matches, report that result and suggest a specific filter adjustment rather than silently broadening the search.

View saved version →

---
name: visibility-gap-analysis
description: Analyze Linkeddit search visibility measurements and recommendation gaps when a user asks why competitors appear, where visibility changed, or what evidence-backed improvements to prioritize.
---

# Visibility gap analysis

Resolve the user's domain with `answer_projects_list`; use returned project IDs. If several projects match, clarify the target before retrieving detailed records. Use the connected tools' current schemas; missing tools are an access limitation, not proof of a particular plan.

For current performance, use `answer_metrics_get`. For a change over time, use `answer_trends_get` with the requested period. Preserve each measurement segment (provider, model, tool version, locale), metric definitions, sample counts, and disclosures. Compare like segments and time windows. `usable=false` or missing observations mean unavailable measurement, not zero visibility. Do not average across segments or attribute a change to a cause without evidence.

For improvement opportunities, call `answer_gaps_list`, then `answer_gap_get` for the gaps most relevant to the user's buyers. Keep recorded observations, cited sources, hypotheses, and suggested actions distinct. A competitor mention alone does not establish a recommendation. Note evidence dates and gaps in coverage.

Return the material findings with source links or returned record IDs, why each matters, and a practical next action. Label your prioritization as analysis rather than a stored score. If the user wants content, draft it in the conversation from the retrieved evidence; don't imply it has been saved or published.

Reading results does not require a new audit. If the user requests a new audit, use the available estimate tool, explain its scope and credit estimate, and obtain agreement before starting it. No automatic project creation, recurring monitoring, or publishing follows from an analysis request.