Pipeline health

Purpose

Produce a dated, evidence-backed view of pipeline coverage, velocity, conversion, concentration, and forecast confidence. The skill calculates what the supplied data supports and leaves policy judgments unknown when the team's definitions are absent.

Usage

• /pipeline-health — review the current confirmed reporting period
• /pipeline-health [period] — review a named month, quarter, or date range
• /pipeline-health forecast — focus on explicit forecast categories and gaps

Evidence, authority, and recovery

Set a report as-of timestamp before reading data. Attach field provenance to every target, deal value, stage, probability, forecast category, date, benchmark, and calculated claim: source path or record ID, source event date, and read as-of time. A file modified time is only a discovery clue, never a substitute for a business event date.

• Use only configured and confirmed stages, targets, probabilities, forecast categories, thresholds, and benchmarks for the requested period. Record the applicable configuration source and effective date. If one is absent, stale, contradictory, or unconfirmed, mark it Unknown; never invent a replacement or import a generic sales convention.
• Missing differs from zero. A blank, unreadable, absent, or conflicting value is Unknown; zero is valid only when the authoritative source explicitly records zero.
• A benchmark needs a source and date. Without one, show the factual metric and Assessment: Unknown — no sourced benchmark.
• Show denominator coverage for every percentage and rate. Verify arithmetic from raw numerators, denominators, stage subtotals, and deal-level contributions.
• Keep analysis read-only. A recommendation is not human authority. Preview any requested change, require explicit confirmation, then read back the result. If a write or read-back fails, report possible partial state, re-read the authoritative record, and wait for fresh human direction.

Method