Agentic Discovery Skill

Purpose

Create discovery and scoping artifacts that turn an agentic development opportunity into verified facts for proposal/SOW, contract, invoice, and delivery work.

System Boundary

This skill owns pre-sale fact-finding and scoping. Hand off to agentic-proposal-skill for buyer-facing scope and SOW documents, agentic-security-review-skill for formal security or launch risk review, agentic-delivery-skill after approval, and agentic-customer-success-skill for ongoing relationship management. Do not use discovery artifacts as final proposals, contracts, invoices, launch signoffs, or public proof.

Core Workflow

1. Identify the discovery need: qualification, intake, stakeholder interview, workflow map, readiness check, risk review, success criteria, evaluation examples, prioritization, recap, or proposal handoff.
2. Gather verified facts only: repeated workflow, inputs, systems, tools, retrieval sources, decision points, approval gates, risks, exclusions, logging, monitoring, documentation, support, and handoff needs.
3. Use references/use-case-decision-table.md to choose the right artifact.
4. Use references/discovery-positioning.md for CompleteTech LLC framing and guardrails.
5. Use references/discovery-catalog.md for near-exhaustive artifact templates.
6. Keep discovery practical and bounded. Do not fabricate client facts, proof, savings metrics, regulated-use assurances, legal claims, or implementation feasibility.

Artifact Selection Guide

• First-pass lead fit: use quick-qualification-checklist.
• New opportunity intake: use client-intake-questionnaire.
• Live discovery call: use workflow-discovery-script.
• Multiple roles or departments: use stakeholder-interview-guide.
• Existing process documentation: use current-state-workflow-map.
• Proposed reviewed-agent process: use future-state-agentic-workflow-map.
• System, document, data, or access review: use systems-and-data-access-checklist.
• API/tool feasibility: use tool-api-readiness-checklist.
• Approval and escalation design: use human-approval-gate-checklist.
• Regulated, sensitive, or risky workflow: use risk-and-excluded-use-checklist.
• Buyer asks what success looks like: use success-criteria-worksheet.
• Need test examples: use evaluation-example-worksheet.
• Decide whether to propose a pilot: use pilot-readiness-scorecard.
• After discovery call: use discovery-recap-memo.
• Handoff into proposal/SOW: use requirements-brief-for-proposal-sow-handoff.
• Missing facts or assumptions: use assumptions-and-unknowns-log.
• Ownership unclear: use stakeholder-responsibility-matrix.
• Dependencies unclear: use implementation-dependency-checklist.
• Sensitive data or retention concerns: use data-sensitivity-and-retention-worksheet.
• Several possible workflows: use workflow-prioritization-matrix.
• Need logging/monitoring requirements: use logging-and-monitoring-needs-worksheet.
• Need support/handoff expectations: use support-and-handoff-discovery-worksheet.
• Need production-readiness check: use pilot-to-production-readiness-checklist.

When several artifacts fit, choose the earliest artifact that resolves the biggest unknown. Do not draft proposal/SOW scope until the workflow, owners, approval points, success criteria, dependencies, and exclusions are clear enough to avoid vague scope.

Quality Rules

• Frame discovery as bounded workflow scoping, not AI brainstorming.
• Always identify human approval points for external communication, production changes, purchases, data export, and material business decisions.
• Capture risks, excluded uses, data sensitivity, and assumptions explicitly.
• Make outputs easy to hand to agentic-proposal-skill.
• Use TBD or open questions for unknowns.

Resource Guide

• references/discovery-positioning.md: load for CompleteTech LLC discovery language and boundaries.
• references/use-case-decision-table.md: load when choosing an artifact.
• references/discovery-lifecycle.md: load for flow from lead qualification through proposal handoff.
• references/discovery-catalog.md: load for the near-exhaustive artifact library.
• references/template-index.json: machine-readable template metadata used by the renderer.
• scripts/render_discovery.py: list discovery artifacts or render a draft with placeholders.

Renderer

python3 scripts/render_discovery.py --list
python3 scripts/render_discovery.py --stage readiness --list
python3 scripts/render_discovery.py --template client-intake-questionnaire --var client_name=Acme --var workflow="support triage"

Rendered artifacts are drafts. Replace placeholders with verified client facts and refine questions for the meeting context.

Rendering to a Branded PDF

Artifacts from this skill are delivered as branded CompleteTech LLC PDF documents, not raw Markdown. The renderer emits the PDF (and prints the Markdown) in one command, using the same reportlab branding engine as the contract skill:

pip install -r requirements.txt
python3 scripts/render_discovery.py --template requirements-brief-for-proposal-sow-handoff \
  --out artifact.pdf --png artifact.png \
  --title "Requirements Brief — Proposal / SOW Handoff" --doc-type "DISCOVERY HANDOFF" \
  --subtitle "Prepared for \x3Cb>Northwind Trading Co.\x3C/b>" --meta "DOCUMENT NO.=DISC-2026-0117" --meta "DATE=2026-05-15" \
  --var client_name="Client Name" --var workflow="support triage"

• --no-pdf emits Markdown only (the original behavior); --no-cover drops the cover page.
• Already drafted the Markdown yourself? Render it directly: python3 scripts/render_pdf.py --markdown artifact.md --out artifact.pdf --logo assets/logo.png --title "...".
• The PDF supports a Markdown subset: #/##/### headings, paragraphs, - bullets, tables, > callouts, **bold**, and [PAGE_BREAK]. PDF requires reportlab; the optional --png preview requires pypdfium2 and pillow. See assets/examples/ for a rendered example.

Certificate Receipt Guidance

The skill remains usable without a classroom key. When certificate credit is needed, use scripts/request_receipt.py after the skill run. The shared class key is provided through CT_CERT_COMPLETION_KEY, --completion-key, or a registry profile; the website claim form receives only the generated receipt code.

Receipt requests include this skill ID: agentic-discovery-skill. The helper sends class/session IDs, the shared key, skill version, generated run ID, optional artifact hash, and metadata to https://cert.complete.tech/api/skill-runs. The student claims the certificate at https://cert.complete.tech/claim with the returned receipt.

Do not print, store, or commit real classroom completion keys.