landscape-discovery-audit

Phase 1: Landscape Discovery & Baseline Audit

You are conducting the discovery phase of a structural overhaul. The substrate-context.md file at the project's working directory tells you WHAT you're auditing and WHY. Your job is to produce a landscape report that the next phase (taxonomy-modeling-design) can consume.

Preconditions

Working directory exists and contains substrate-context.md.
If substrate-context.md is missing, stop and ask the user to run the pentaphase-orchestrator Phase 0 first, OR have them dictate the substrate, driving force, primary failure point, operational roles, and time horizon so you can write substrate-context.md yourself.

Read substrate-context.md in full before starting any work stream.

Four work streams

The four streams produce four named outputs that you will combine into the final report. Run them in any order; produce each as its own section in the working directory before composing the report.

Stream 1 — Inventory Assets

Document every existing core output, component, and data point in the substrate.

For each asset, capture:

Field	Required	Notes
Name	yes	Canonical name in the system, plus any aliases
Location	yes	Absolute path, URL, registry id, or coordinates
Owner	yes	Person, team, or role responsible
State	yes	active / archived / deprecated / orphaned / unknown
Last touched	yes	ISO date if discoverable; otherwise "unknown"
Size / scale	yes	rows, files, bytes, or domain-native unit
Dependencies	optional	Other assets this one depends on
Notes	optional	Anything anomalous worth flagging

Output as a markdown table or sectioned list. Descend until you have leaves, not branches — "the docs folder" is a branch; the individual files inside are leaves.

If the substrate is too large to enumerate exhaustively in one pass, sample by category and declare the sampling method explicitly.

Stream 2 — Map Current Flow

Track how items travel from creation to final archive. For each distinct flow:

Trigger — what initiates the flow (event, schedule, human action)
Stages — ordered list of states the item passes through
Actors — who or what advances each stage (human role, automation, hybrid)
Handoffs — where state passes between actors (these are the friction-prone points)
Termination — how the flow ends (archive, deletion, escalation, abandonment)
Volume — items per unit time (or backlog depth if flow is irregular)

If multiple distinct flows exist, document each separately. Number them. Use prose, ASCII diagrams, or sequence lists — whichever is clearest.

Stream 3 — Identify Friction

Pinpoint structural bottlenecks, communication gaps, and systemic inefficiencies. For each:

Friction point — specific, named, located in the flow
Type — bottleneck / gap / inefficiency / drift / silo / coordination cost / quality leak
Frequency — how often it bites (every event, weekly, monthly, sporadic)
Impact — cost in time, money, quality, morale, opportunity
Root cause hypothesis — your best read on the underlying mechanism

Resist generic complaints. "Slow process" is not a friction point. "Tickets sit in the 'awaiting-review' state for an average of 6 days because reviewers are not paged" is a friction point.

Stream 4 — Define Value Metrics

Establish clear success indicators for speed, accuracy, and utilization. At minimum, one metric per dimension. For each:

Name
Definition — what is measured, how, with what tool
Current baseline — current value with measurement method and date
Target — where we need to be after the overhaul
Cadence — how often it's measured
Threshold — value at which an alert/escalation fires (optional but recommended)

Metrics must be measurable. "Better UX" is not a metric. "Time from ticket creation to first response, p50, in minutes" is.

Composing phase-1-landscape-report.md

Combine the four streams into a single file at <working-dir>/phase-1-landscape-report.md. Structure:

# Phase 1 — Landscape Report

**Substrate:** <name from substrate-context.md>
**Date:** YYYY-MM-DD
**Preconditions:** Read substrate-context.md (path)
**Postconditions:** Ready for Phase 2 (taxonomy-modeling-design)

## 1. Executive summary

3–5 sentences on the substrate's current shape and biggest pain.

## 2. Assets inventory

[full table or sectioned list]

## 3. Current flow map

[per-flow sections]

## 4. Friction register

[per-friction-point list]

## 5. Value metrics

[per-metric definitions with baselines]

## 6. Open questions for Phase 2

[explicit list of gaps, ambiguities, or decisions deferred to taxonomy design]

Gate criteria (auditor will check)

The report passes Phase 1's gate iff:

Asset coverage ≥ 80% — no major asset class missing. If sampling was used, the sampling method is declared.
At least one flow mapped end-to-end — from trigger to termination, with stages, actors, and handoffs.
Friction register has ≥ 3 named points — each with a type, a frequency, an impact, and a root cause hypothesis.
Value metrics cover speed AND accuracy dimensions — each with a current baseline and a target. Utilization is encouraged but not required.
Open questions for Phase 2 section is present — even if empty (an explicit "no open questions" is acceptable; an absent section is not).

If you cannot meet a gate criterion, flag the gap explicitly in the "Open questions" section and surface it to the user before invoking the next phase.

Anti-patterns

Don't invent assets you can't verify. If something might exist, mark it "suspected — needs confirmation" rather than asserting it.
Don't smooth over confusion. If two roles claim ownership of the same asset, document the conflict — that IS the friction point.
Don't skip flow mapping by saying "it's complicated". If it's complicated, that's a Phase 2 problem to solve. Map what's there.
Don't propose solutions in this phase. Solutions belong in Phase 2 (taxonomy) or Phase 3 (environment). Phase 1 is descriptive only.

About this skill