convert-skill

Custom Skill Converter

Take any custom Claude skill, prompt, or workflow and rewrite it as a Raylu-optimized version that uses our MCP tools. Walk through these steps ONE AT A TIME. Do not skip steps.

Step A: Ingest the Source Skill

Ask: "Paste the skill, prompt, or workflow you want to convert. It can be a Markdown file, a system prompt, a set of instructions, or rough notes — anything that describes what you want Claude to do."

Read the full text carefully. Do NOT skim. Identify:

• The skill's name and primary objective
• The trigger conditions ("use when...")
• The discrete steps or stages
• The questions the skill asks the user along the way
• The external tools, websites, APIs, or manual processes the skill currently calls (Crunchbase, LinkedIn, Pitchbook, Google search, spreadsheets, etc.)
• Any output formats, tables, or artifacts the skill produces

Present a brief summary back to the user before doing anything else:

"Here's what I understood from your skill: {objective}. It has {N} steps. It currently uses {list of external tools or manual processes}. Confirm before I convert it, or correct anything I got wrong."

Step B: Map External Actions to Raylu Tools

For each step in the source skill, identify which Raylu MCP tools can replace or enhance it. The full Raylu tool catalog is below — use these as the building blocks:

Company data & lookup

• search_companies — disambiguate by name across the warehouse
• get_company / get_companies — full firmographic profile (single or bulk)
• research_company — research and start tracking a company
• list_companies — list all saved companies

Firmographic search

• firmographic_search — structured search with filters: headcount, country, state, city, last_funding_stage, last_funding_amount, total_funding_amount, financing_status, founded_date, revenue, monthly_visits, industries, keywords, all_investors, all_funding_leads

Market maps

• start_market_map, update_market_map, get_market_map, get_market_map_data, list_market_maps
• create_ai_enrich_column, enrich_market_map — add and run AI enrichment columns
• create_ma_map — build an M&A-anchored map in one operation
• add_company_to_market_map

Lists

• create_list, get_list, list_lists, add_company_to_list

Contacts & outreach

• find_company_contact — verified email + phone for the primary contact
• list_strategies, create_strategy — outreach cadences
• generate_campaign — generate a full sequence
• start_campaign, pause_campaign, cancel_campaign, complete_campaign_step
• update_email, rewrite_email — edit individual emails
• get_campaign, list_campaigns
• get_outreach_schedule, get_daily_send_count

Other

• setup_sourcing_routine, setup_ma_mapping — bootstrap recurring workflows
• get_run_status

For each source step, output a mapping in this table format:

Source step	Raylu tool	What changes
"Look up the company on Crunchbase"	get_company	Pulls Raylu's enriched profile instead of opening Crunchbase manually
"Find a contact on LinkedIn"	find_company_contact	Returns verified email + phone in one call
"Filter to Series A SaaS companies"	firmographic_search	Structured filter on last_funding_stage and keywords

Present the full mapping table to the user. Explicitly flag any source steps that have NO Raylu equivalent — those should stay as-is or be handled outside the tool chain.

Step C: Identify Optimization Opportunities

Look beyond 1-to-1 replacement. Suggest improvements the original skill couldn't access:

• Can a market map replace a multi-step research loop?
• Can an AI enrichment column automate something the user was doing manually per company?
• Can research_company + find_company_contact + generate_campaign be chained instead of run separately?
• Can a scheduled routine (setup_sourcing_routine) replace something the user runs ad-hoc?
• Can a saved list (create_list + add_company_to_list) replace a spreadsheet?

Present the optimizations and ask the user which to apply. Do NOT silently change the skill's intent — flag every meaningful structural change before making it.

Step D: Rewrite in Raylu Workflow Style

Generate the new skill using this exact structure (matches the existing Raylu workflows):

1. Title — preserve or refine the original skill's name
2. One-line objective — what the skill does, written as an imperative
3. Top-level rule — e.g., "Walk through these steps ONE AT A TIME. Do not skip steps."
4. Optional autonomy rule — if Claude should act without confirmation in certain cases, state it up front
5. Steps — ## Step A, ## Step B, etc. Each step must:
   • Lead with the user-facing question (if any), formatted as Ask: "..."
   • Name the exact Raylu tool calls in order
   • Explain what to do with the output and when to confirm with the user
6. Output format — if the original produced a table, preserve the column schema. If it produced a recommendation, preserve the framing.

Tone match: direct, imperative, no fluff. Read M&A Mapping, Sourcing Routine Setup, and Outreach Sequence Builder for reference — match their cadence and structure.

Step E: Validate and Hand Off

Present the converted skill to the user inside a fenced code block (so it copy-pastes cleanly). Then ask:

1. "Does this preserve the intent of your original skill?"
2. "Any steps you'd like to drop, reorder, or rename?"
3. "Want me to save this as a Markdown file you can share with your team or add to the Raylu MCP Workflows library?"

If the user approves, format the final version as Markdown with the same conventions used by the workflows in the Raylu MCP Workflows library.