Auditing A Pigment Application

Auditing a Pigment Application

Run a structured, read-only audit. Do not modify anything unless the user explicitly asks for fixes after reviewing the report.

CRITICAL -- Collect Formula Health Evidence First

Start every audit with tool:list_issues to surface active formula computation errors. These are highest-severity because they produce wrong or missing outputs.

For each errored block, retrieve its formula with tool:search_metrics_and_lists (show_details: true) and identify the root cause (type mismatch, missing reference, scenario-specific failure). Use tool:get_metric_dependencies on errored blocks to trace downstream impact and distinguish root causes from cascading symptoms.

CRITICAL -- Trace Dependency Fragility

Use tool:get_metric_dependencies on critical output metrics (cashflow totals, P&L summaries, KPIs).

Flag:

• Long chains where a single upstream error cascades to many outputs
• Missing intermediate metrics (too much logic in one formula)
• Fragile roll-forward patterns (e.g. CUMULATE + SELECT: Month-1 without clear beginning/ending balance separation)

HIGH -- Scan Formula Patterns

Use tool:search_metrics_and_lists with formula_regex_search pattern queries. Do not rewrite formulas during the audit. For the full anti-pattern reference, see skill:writing-performant-formulas. For AR anti-patterns, see skill:securing-with-access-rights.

Flag these patterns (highest priority first):

• Unguarded ACCESSRIGHTS: no IFDEFINED('Users roles', ...) guard — AR computed for every user
• ACCESSRIGHTS guarded only by IFDEFINED(User, ...): flag for review — not a substitute for 'Users roles'
• ACCESSRIGHTS(x, FALSE): densifies denied cells; should use BLANK
• ISBLANK / ISNOTBLANK: densifies; should use ISDEFINED / IFDEFINED
• [ADD: X][FILTER: ...]: densify-then-subset; use IF(condition, value)
• [FILTER: NOT ...]: use [EXCLUDE: ...]
• Heavy PREVIOUS chains / multiple PREVIOUS(...) calls: check if CUMULATE, FILLFORWARD, or SELECT fits
• IFBLANK(X, PREVIOUS(...)) or IFBLANK(X, PREVIOUSOF(...)): use FILLFORWARD
• [ADD: X][REMOVE: X] / [REMOVE: X][ADD: Y] round-trips: use BY with mapping
• Chained [BY:][BY:] on Transaction List: drops properties; use single BY with comma-separated mappings
• [REMOVE: Version]: collapses scenario meaning; verify intentional
• REMOVE in formulas: scope loss; check with tool:get_metric_dependencies
• 2-arg IF(cond, expr) on 6+ dim metrics with non-trivial second arg: use expr[FILTER: cond]
• Hard-coded year / DATE(YYYY, ...): use Date-typed or Dimension-typed input metrics
• Periods or colons in block names: breaks formula references
• Unresolved Copy of X default names: leftover from duplication
• Formula metrics with input overrides enabled: audit risk if untracked

HIGH -- Check Cycles and Versions

Use tool:list_cycles to check planning cycles. Use tool:search_metrics_and_lists with kind: ["Dimension"] and a name regex to verify a Version dimension exists. A missing Version dimension is a high-severity structural gap; see skill:building-versions-and-planning-cycles for fixes.

HIGH -- Verify Calendar Coverage

Use tool:calendar_get to read the current configuration. Flag:

• Calendar end date before the next fiscal year (blocks future planning)
• Missing Quarter dimension when the app is FP&A-oriented
• Daily calendar with a large range (performance risk for iterative formulas)

For fixes, see skill:setting-up-calendar.

MEDIUM -- Review Scenario Governance

Use tool:list_scenarios. Flag:

• Duplicate or near-duplicate names (case differences like Budget vs budget)
• Non-descriptive names (child, test, scenario1)
• Total count above 5-6 in a non-enterprise app (each scenario increases recalculation cost)
• Scenarios with formula errors (cross-reference with formula health findings)

MEDIUM -- Audit Library and Shared Content

Use tool:list_application_libraries. Flag:

• Libraries enabled but providing no actively used blocks
• Old or versioned library names (e.g. FY24 library on an FY25 app)
• Excessive shared surface exposing internal blocks
• Shared metrics consumed locally but not through the library mechanism (direct cross-app references instead of PULL/PUSH)

MEDIUM -- Check Modeling Governance

Use tool:search_metrics_and_lists and tool:get_metric_dependencies to identify structural debt:

• Repeated logic across metrics: Same subexpression duplicated in multiple formulas instead of centralized in a shared intermediary metric
• Multi-purpose metrics: A single metric doing aggregation, filtering, and business logic; should be split into focused layers
• Hidden-metric access gaps: Metrics with AR Apply rules that reference dimensions not in the metric's own structure, or metrics consuming AR-protected data without their own AR coverage
• Access rights enforced only via Board visibility: Data is hidden on boards but has no AR metric backing it; users with Block Explorer access can see everything. See skill:securing-with-access-rights

LOW -- Inspect Naming Hygiene

Use tool:search_metrics_and_lists with friendly_name_regex_search to find blocks matching test/copy/temporary patterns:

• copy, Copy, COPY: Leftover duplicates
• test, Test, TEST: Experimental blocks not cleaned up
• new metric, New Metric, NewBlock: Default names never renamed
• tmp, TMP, ZZ, TBD: Explicitly temporary blocks
• Hex-dash sequences (UUID-like): Auto-generated names

Count them, note their folder, flag as cleanup candidates.

LOW -- Find Unused Blocks

Use tool:search_metrics_and_lists with kind: ["Metric"] to page through all metrics, then tool:get_metric_dependencies on each to find orphans (zero usages). Batch dependency checks in parallel.

Safe to delete: no formula references, no views, no boards, no automations, no ARM. Not safe: displayed on a board/view, shared via library (other apps may consume), input metrics with user data, metrics with overrides, or settings/parameters folder items. Always present the list to the user before deleting; group by folder; highlight input and shared metrics separately.

LOW -- Check Folder Organization

Use tool:search_metrics_and_lists with parent_folder_regex_search: ["^/$"] or an empty-path filter to identify blocks in root / no folder. Check whether a standard folder structure exists (numbered prefixes like 0. Settings, 1. Dimensions). Mixed-purpose or flat layouts are a maintenance risk. For fixes, see skill:naming-and-organizing-applications.

LOW -- Evaluate Board Quality

Use tool:search_boards and tool:get_board for suspicious boards. Flag:

• Total board count above 20
• UUID-like or meaningless board names
• Duplicate boards (same or very similar names)
• Homepage action cards pointing to wrong destinations
• Boards with more than 20 widgets (load-time risk)

Report Structure

1. Executive summary: 2-3 sentences on model health and top priorities
2. Findings by severity: CRITICAL (active errors) → HIGH (structural risks) → MEDIUM (governance debt) → LOW (cosmetic). Name specific blocks with mention: syntax.
3. Remediation priority: numbered list; errors first, then structural, then hygiene
4. Next steps: offer cleanup plan, fixes, or deeper performance analysis

Audit Checklist

Before delivering the report, confirm all checks ran:

• Formula health (tool:list_issues + dependency tracing)
• Dependency fragility (critical output chains)
• Formula patterns (sparsity, ACCESSRIGHTS guards, iterative, structural, dimension-aware, naming)
• Cycles and versions (tool:list_cycles)
• Calendar coverage (tool:calendar_get)
• Scenario governance (tool:list_scenarios)
• Library surface (tool:list_application_libraries)
• Naming hygiene (test/copy/tmp/UUID patterns)
• Unused blocks (tool:get_metric_dependencies zero usages)
• Modeling governance (repeated logic, multi-purpose metrics, hidden-metric access gaps)
• Folder organization (blocks in root)
• Board quality (tool:get_board)