/mesh-tools

Activate the socialmesh skill. No argument required.

What this does

Reads every data/harnesses/<platform>/latest.json and prints the tool definitions — name, description, input schema, and the analysis prompt that embeds the profile context.

Steps

1. Enumerate harnesses

find data/harnesses -name "latest.json" 2>/dev/null | sort

If no harnesses exist, tell the user to run /mesh-scan <platform> to generate them and stop.

2. For each harness, extract tools

Read the tools array from each harness file. For each tool, extract:

• name
• description
• input_schema (summarized, not the full JSON)
• analysis_prompt (first 200 characters with ellipsis if longer)

Also read:

• platform
• generated_at
• source_profile_synthesized_at

3. Print grouped output

Group by platform. For each platform:

Platform: instagram
Harness generated: <timestamp> (from profile: <timestamp>)

  analyze_feed_bias
    Analyzes the political and topical lean of the feed.
    Inputs: none
    Prompt: "Given this Instagram profile with lean_score=0.3 and dominant topics..."

  identify_echo_chamber
    Estimates the degree of feed homogeneity and echo risk.
    Inputs: none
    Prompt: "This feed has an echo_index of 0.71, driven by..."

  recommend_doors
    Suggests accounts that would introduce underrepresented viewpoints.
    Inputs: { "count": number (default 3) }
    Prompt: "The following topics are underrepresented in this feed..."

  explain_top_authors
    Explains which authors dominate the feed and why.
    Inputs: { "limit": number (default 10) }
    Prompt: "The top authors by feed weight are..."

  summarize_topic_mix
    Breaks down the topic distribution in the feed.
    Inputs: none
    Prompt: "Topic distribution for this feed: cooking 22%, travel 18%..."

4. Print summary

After all platforms, print:

• Total tool count across all platforms
• Platforms with no harness: list them and suggest running /mesh-scan <platform>
• Note: "These tools are used automatically by /mesh-bias. You can also pass them to any AI agent via the harness system_prompt."