Fullstory Org Discovery

Fullstory Org Discovery

The goal is to produce a concise, accurate context file that captures the signal from a Fullstory org in a form that's immediately useful to both humans and analysis skills. Quality matters more than completeness — a curated file of 20 meaningful entries is far more useful than a raw dump of 200 auto-learned pages.

Step 1 — Identify the org

If the org name or slug hasn't been provided, ask before proceeding:

"What's the org name or slug? I'll save the context file to ~/claude/accounts/{slug}-fullstory-context.md."

Once you have a name, derive a slug (lowercase, hyphens, no spaces) and the output path.

Step 2 — Parallel discovery

Fire all four of these simultaneously — they're independent and waiting is wasteful:

1. get_managed_funnels() — all defined funnels
2. get_pages(regex=null) — full page list (you'll curate this in Step 3)
3. discover_org_context with these queries and limit=5:

   checkout, purchase, order, cart, payment,
   login, signup, register,
   search, filter,
   confirm, success, complete, thank you,
   error, failed,
   product, listing, item,
   conversion, revenue
   

4. get_opportunities(time_range="30d") — StoryAI's pre-ranked issues org-wide

Step 3 — Curate pages

The raw page list may contain hundreds of auto-learned entries. Keep:

• All user-defined pages — these are always meaningful, whatever they're named
• Auto-learned pages whose name or friendly_id contains any of: checkout, cart, product, search, login, register, sign, confirm, order, home, landing, account, error, 404, basket, booking, payment, success, thank

Discard: root-only paths (/), UUID/hash paths, pages with no recognisable business purpose.

The goal is a curated list that a trading manager or COO could read and immediately understand — not an exhaustive inventory.

Step 3b — Validate events against key pages

Pick the top pages by funnel importance — include all pages that appear in funnels, then fill remaining slots with high-signal pages (home, search results, PDP) up to a maximum of 10. Call discover_groups for each in parallel:

discover_groups(page_id=X, relative_time_range="30d", limit=5)

Use the results as a validation and enrichment pass — not a new output section:

• Confirm active events: Cross-reference the groups returned against the Defined Events and Named Elements you already found. Events/elements that appear in page signals are confirmed as genuinely active — note this in their table row (e.g. add a "Pages" column or parenthetical).
• Flag outliers: If a group appears prominently in page signals but has no match in your discovered events/elements, it's a potential gap in the org's instrumentation or a signal your discovery queries missed. Add it to the Notes section as: "Unmatched signal on [page]: [element/event] — consider including in analysis."

If everything from discover_groups matches what you already found, no output change needed. Only surface what's genuinely new or confirmatory.

Step 4 — Synthesise and characterise

Now reason over everything you've collected. The output should make sense to a non-technical senior stakeholder, not just a developer:

• Infer site type: eCommerce, SaaS, financial services, travel, media, etc. — use the vocabulary of funnels, pages, and events to make this call
• Map funnels to business purpose: "Checkout Funnel" → "The primary path from product selection to completed purchase"
• Write a site characterisation: one paragraph a COO could read. What does this site do? What are the key user journeys? What's the scale of implementation (shallow/rich)?
• Flag implementation gaps: if there are no funnels, no named events, or very few user-defined pages, note this. It means analysis will be less precise and is important context for a POC.

Step 5 — Write the context file

Write to ~/claude/accounts/{slug}-fullstory-context.md using this structure exactly:

# Fullstory Org Context: {Org Name}
Last updated: {YYYY-MM-DD}
Org slug: {slug}

## Site Characterisation
**Type:** {eCommerce / SaaS / financial services / travel / other}

{One paragraph describing the site: what it does, key user journeys, 
and the richness of the Fullstory implementation.}

## Funnels
| Name | ID | Business Purpose |
|------|----|-----------------|
| ... | ... | ... |

## Key Pages
| Name | ID | Type | Business Purpose |
|------|----|------|-----------------|
| ... | ... | user-defined / learned | ... |

## Defined Events
| Name | ID | Meaning |
|------|----|---------|
| ... | ... | ... |

## Named Elements
| Name | ID | Purpose |
|------|----|---------|
| ... | ... | ... |

## Custom User Variables
{List any user properties defined in this org — useful for building segments.
If none, write: "None defined."}

## Top StoryAI Opportunities (last 30 days)
| Title | Severity | Users Affected | Metric ID | Group ID |
|-------|----------|---------------|-----------|----------|
| ... | ... | ...% | ... | ... |

{Keep Metric ID and Group ID — analysis skills need these for drill-in.}

## Notes for Analysis Skills
{Any flags that would affect downstream analysis:
- No funnels defined → conversion analysis will need manual scoping
- No named events → rely on page/click patterns only
- Shallow implementation → StoryAI opportunities may be limited
- Unusual naming conventions → note them here
- Any other context useful to an analysis running against this org}

If a section has no data (e.g. no named elements), write "None found" rather than omitting the section — analysis skills need to know what's absent, not just what's present.

Step 6 — Report to user

After writing the file, you MUST print this summary to the user. This is the primary output they see — don't skip it or bury it in prose.

Org discovery complete — {Org Name}
Saved to: ~/claude/accounts/{slug}-fullstory-context.md

Site type: {inferred type}
Found: {N} funnels · {N} key pages · {N} defined events · {N} named elements

Top 3 issues (StoryAI):
1. {title} — {severity}
2. {title} — {severity}
3. {title} — {severity}

Notes: {any flags from implementation gaps or outliers, or "None — implementation looks well-structured."}

Example:

Org discovery complete — Cargo Rentals
Saved to: ~/claude/accounts/cargo-rentals-fullstory-context.md

Site type: Peer-to-peer car rental marketplace
Found: 15 funnels · 23 key pages · 34 defined events · 26 named elements

Top 3 issues (StoryAI):
1. Script errors on checkout — 133 users (INVESTIGATING)
2. Dead clicks on "Add to Cart" button — 41 users (TICKET_OPENED)
3. "No cars found" modal after city search — 31 users (TICKET_OPENED)

Notes: Duplicate Revenue Event defined — verify KSNijoJTTEpe is the active one before building funnels. No custom user variables found.