Skills & agents

The Plugin ships more than the connector. On top of the 32 tools it adds skills — the procedures that know how to build, edit and audit a Genudo pipeline — and agents, specialists that run a whole job end to end.

Paste any prompt below into Claude, ChatGPT or Codex to try it. Every write is previewed as a diff and confirmed before it touches your account.

	Claude plugin	ChatGPT plugin
Tools	32	32
Skills	22	28
Agents	6	—

Connector only?

If you connected by URL instead of installing the plugin, you have the tools but none of the skills or agents. See Connect a client.

Build a pipeline

Write · From a business goal to a live agent — interview, draft, then create it.

Skill	What it does
discover-pipeline-requirements	Interviews you for everything a new agent needs: goal, channel, offer, stages.
author-pipeline-brain	Drafts the persona and the global instructions — who the agent is and how it behaves.
author-stages	Designs each stage: entry condition, in-stage flow, data to collect, won/lost nature.
provision-pipeline	Creates the pipeline, stages, variables and actions in your account from the draft.
configure-pipeline-settings	Name, cost and message ceilings per conversation, follow-up limits, knowledge options.

Try these prompts

Build me a WhatsApp sales agent for a summer camp — collect name, phone, branch and child age, and escalate special-needs questions.

Interview me for a new support pipeline, then show me the stages before creating anything.

Edit a live agent

Write · Change what a running agent says — safely, with a diff you approve first.

Skill	What it does
edit-pipeline-instructions	Edits persona, global instructions, stage instructions or entry conditions.
instruction-guides	The authoring rules and templates — read before writing any instruction text.
editing-playbook	The safe workflow: load, mirror, stage the edit, diff, confirm, push, record.
configure-ai-model	Picks the AI model, or a routed pool with router/simple/moderate/complex tiers.
configure-stage-opening-message	The message a stage sends when an opportunity enters it — AI-written or fixed.

Try these prompts

Make the agent stop promising delivery dates, and show me the diff before pushing.

Rewrite the qualification stage so it asks for budget before booking a call.

Automate & integrate

Write · Send data out to your CRM, Sheets, Zapier, Make or n8n at the right moment.

Skill	What it does
manage-pipeline-variables	The variables actions send: system fields, AI-extracted values, action outputs, constants.
build-action-webhook	Builds the outbound HTTP action — method, URL, headers, payload mapped to variables.
wire-stage-automation	The whole loop: collect the data, fire the action when ready, tell the customer.
design-stage-followups	Timed follow-up sequences per stage, for when a lead goes quiet.

Try these prompts

When name, phone and course are collected in the closing stage, register the lead to Google Sheets and tell them someone will reach out.

Add a follow-up ten hours after the lead goes quiet in the qualification stage.

Knowledge & opportunities

Read + Write · What the agent knows, and the leads it is working.

Skill	What it does
manage-knowledge-base	Knowledge tables the agent searches at runtime — create them, fill them, verify retrieval.
manage-opportunities	Bulk updates across leads: status, stage, priority, notes, tags.

Try these prompts

Add our pricing to the agent's knowledge: Pro is $49/month with priority support.

Move every opportunity tagged 'no-budget' in the demo stage to lost, and note why.

Diagnose & analyse

Why the agent said that, where leads drop off, what it all costs.

Skill	What it does
diagnose-agent-behavior	Finds the instruction, stage or action responsible for a misbehaviour, and the fix.
analyze-conversation	Reads a real conversation turn by turn and explains what went wrong, and where.
analyze-funnel	Opportunity counts and distribution across stages, statuses, priorities and tags.
workspace-report	Account totals, AI cost and response time, messaging volume by provider and period.

Try these prompts

Why did my agent say that? Here's the conversation link — find the cause and propose a fix.

How's my funnel doing this month, and where are leads dropping off?

Migrate an old pipeline

Write · Bring a pipeline written the old way onto the current authoring approach.

Skill	What it does
migrate-intake-legacy	Captures the legacy persona, instructions and stage structure into a structured spec.
migrate-map-to-new	Rewrites that spec against today's templates, with a report of what moved and what dropped.

Try this prompt

This pipeline uses the old instruction format — capture it, then rebuild it the current way and show me what changed.

Agents

Autonomous · Specialists that run a whole job end to end. They run in Cowork and Claude Code.

Agent	What it does
pipeline-architect	Designs and builds a new pipeline from a business goal, start to finish.
pipeline-doctor	Diagnoses a misbehaving agent, root-causes it and applies the fix.
automation-engineer	Builds the integrations and the variables that feed them.
revenue-analyst	Funnel health, conversation quality, AI cost — read-focused, confirms before changing.
kb-librarian	Curates the knowledge base when the agent answers with wrong or outdated facts.
pipeline-migrator	Takes an old-format pipeline and rebuilds it against the current approach.

Try these prompts

Use the pipeline-architect agent to build a Messenger pipeline for my clinic.

Use the revenue-analyst agent to review last month and tell me what to fix first.

---

Worked examples

Real jobs, and the parts that do the work — the agent orchestrates, the skill supplies the procedure, the tool makes the change.

Launch a new sales agent

Interviewed for requirements, stages drafted for review, then created in your account.

You paste

Build me a WhatsApp sales agent for a summer camp, and show me the stages before you create anything.

Then it runs

#		Kind	What happens
1	pipeline-architect	agent	Takes the job and runs the sequence below, reporting back at each checkpoint.
2	discover-pipeline-requirements	skill	Interviews you: business goal, channel, what you sell, what to collect, when to escalate.
3	author-pipeline-brain · author-stages	skill	Drafts the persona and every stage as local files you read and correct before anything ships.
4	create_pipeline · create_stage	tool	Creates it in your account — only after you say go. Nothing before this step touched it.

Fix an agent that went off-script

Reads the real conversation, traces it to the instruction at fault, edits with a diff you approve.

You paste

This conversation went wrong — find the instruction responsible and fix it, with a diff first.

Then it runs

#		Kind	What happens
1	pipeline-doctor	agent	Picks up the complaint and works backwards from the conversation you gave it.
2	list_messages	tool	Pulls the real transcript — the actual turns, not a summary of them.
3	analyze-conversation	skill	Reads it turn by turn and pins the moment it drifted to the instruction that caused it.
4	editing-playbook · edit-pipeline-instructions	skill	Loads the current text and stages the smallest edit that fixes it, then shows you a before/after diff.
5	update_stage	tool	Pushes the corrected stage once you approve the diff.

Register leads into your CRM

Creates the variables the payload needs, then the webhook, then wires it to the right stage.

You paste

When a lead is qualified, send name, phone and interest to our n8n webhook and confirm to the customer.

Then it runs

#		Kind	What happens
1	automation-engineer	agent	Works out what the integration needs before building any of it.
2	manage-pipeline-variables	skill	Decides which variables the payload needs — contact fields, AI-extracted values, constants.
3	create_variable	tool	Creates the missing ones, because an action cannot reference a variable that does not exist yet.
4	build-action-webhook	skill	Builds the webhook: method, URL, headers, and the payload mapped to those variables.
5	create_action · update_stage	tool	Attaches it to the stage and sets when it fires and what the customer is told afterwards.

Report on last month

Funnel distribution, AI spend and response times, pulled and read back as findings.

You paste

Give me last month's funnel report, AI cost per deal, and the top three things to improve.

Then it runs

#		Kind	What happens
1	revenue-analyst	agent	Gathers the numbers first, then interprets them — read-only throughout.
2	get_account_summary · get_ai_performance	tool	Pulls account totals, AI spend and response times for the period.
3	analyze-funnel	skill	Adds opportunity distribution across stages to show where leads actually stall.
4	workspace-report	skill	Reads it back as findings and ranked fixes, not a table you still have to interpret.

For the full parameter reference of every tool named above, see Tools reference.