IntentSignals Skill

Finds distribution opportunities and validates product ideas by scanning online communities. Takes a natural language description of what you have, determines the mode, scans platforms via last30days, and outputs actionable recommendations.

When to Use

• User wants to find places to share a blog post, article, or content
• User wants to validate if a product idea has demand
• User wants to know what questions people ask about a topic before writing
• User wants to find communities where their expertise would be valued

Don't use for: monitoring brand mentions (use last30days directly), competitor tracking, or real-time alerts.

Prerequisites

• last30days skill installed and functional
• At least one source configured (Reddit works without API keys)

Mode Detection

Parse the user's input to determine which mode to use:

Input Pattern	Mode	What to Scan For
"I wrote / I have / here's my post" + URL or title	Content Distribution	Active threads where this content answers a question
"I'm thinking of building / I want to create"	Product Validation	Existing pain points, competitors, feature requests
"I want to write about / what should I write"	Content Validation	Top questions, knowledge gaps, real phrases people use
"Where should I share / promote / distribute"	Content Distribution	Active communities where the content adds value
"Is there demand for / do people need"	Product Validation	Complaints about existing solutions, unmet needs

If ambiguous, ask: "Are you looking to validate an idea, plan content, or distribute something you've already written?"

Procedure

Step 1: Parse Intent and Extract Keywords

From the user's input, extract:

• Core topic: What is this about in 3-5 words?
• Audience: Who would care about this? (e.g., CTOs, indie makers, developers)
• Platform hints: Any platform the user mentioned?
• Mode: Determined from the table above.

Step 2: Generate Search Queries

Based on the mode, generate 3-5 search queries that capture how REAL PEOPLE talk about this topic (not how the user describes their product).

Content Distribution mode: Generate queries like:

• "looking for advice on [topic]"
• "how do you handle [problem]"
• "recommendations for [category]"
• "struggling with [pain point]"

Product Validation mode: Generate queries like:

• "frustrated with [existing solution]"
• "wish there was a tool for [need]"
• "alternative to [competitor]"
• "does anyone know [capability]"

Content Validation mode: Generate queries like:

• "what is [topic]"
• "how to [do thing related to topic]"
• "best practices for [topic]"
• "mistakes to avoid in [topic]"

Step 3: Scan Platforms via last30days

For each search query, use the last30days skill:

/last30days [query]

Scan at minimum:

• Reddit (always — richest intent signals)
• Hacker News (for tech/product/startup topics)
• X/Twitter (for real-time discussion)

Optionally scan:

• YouTube (for video discussions)
• Podcasts (for long-form mentions)
• ProductHunt (for product-related topics)

Collect results from all queries. Deduplicate by URL.

Step 4: Score and Rank Opportunities

For each result, assess:

1. Relevance (1-10): How closely does this match the user's content/idea?

2. Intent signals (1-10): Is the poster actively seeking help/advice, or just discussing?

3. Engagement potential (1-10): Is the thread active? Will a comment be seen?

4. Spam risk (1-10): How natural would it be to mention the user's content here?

Ranking formula:

score = (relevance  0.4) + (intent  0.3) + (engagement  0.2) + ((10 - spam_risk)  0.1)

Select the top 5 opportunities.

Step 5: Generate Actionable Output

For each of the top 5 opportunities, provide:

### [N]. Thread/Post Title
- Platform: [Reddit r/X / HN / X]
- Why it fits: [1-2 sentences explaining the match]
- Suggested approach: [What to say — lead with value, not a link]
- Spam risk: [N]/10
- Timing: [Post now / within 24 hours / wait for more engagement]

Suggested approach rules:

• Lead with an insight or answer, THEN mention the content naturally
• Never open with a link
• On Reddit: comment as a helpful community member, not a promoter
• On HN: be technical and opinionated, reference specific points
• On X: ask a question or share a data point, thread if needed

Step 6: Log the Run

Append to wiki/produce/hermes/products/intentsignals/tracking/daily-log.md:

### Run [N]
- Time: [HH:MM]
- Input: "[user's input]"
- Mode detected: [mode]
- Platforms scanned: [list]
- Opportunities found: [count]
- Selected: [which ones the user picked]
- Action taken: [what was posted]
- Spam risk: [N]/10
- Result: pending

Output Format

Always structure the output as:

# IntentSignals Report

Mode: [Content Distribution / Product Validation / Content Validation]
Input: "[user's input]"
Scanned: [platforms] (last 30 days)
Date: [today]



Top 5 Opportunities

1. Title
...

2. Title
...



Quick Actions
- Best opportunity: #N — [one-line reason]
- Suggested first move: [specific action]

Pitfalls

• Don't keyword-stuff. The user's product description and the search queries should be different. People don't search for your product name — they search for their problem.
• Don't recommend dead threads. Skip anything with < 5 comments or > 30 days old (for distribution mode).
• Don't spam. If the top 5 all have spam risk > 5, say so: "These threads are risky — consider waiting for better opportunities."
• last30days may fail. Reddit scraping is fragile. If it fails, fall back to web_search for the same queries and note the degraded quality.
• Log even failures. If a run produces nothing useful, log it — that's data too.

Verification

After a run, confirm:

• [ ] Mode was correctly detected
• [ ] At least 3 opportunities were found (or explain why fewer)
• [ ] Each opportunity has a suggested approach (not just a link)
• [ ] Spam risk is assessed for each
• [ ] Run is logged to daily-log.md