Skip to content

JasonColapietro/suede-creator-skills / skills/suede-ops-assessment

suede-ops-assessment

Verified

Suede Operations Assessment Skills: map real work, silos, friction, ownership, costs, and adoption signals from operator evidence. Use when asked "where is our team losing time" or to rank operational opportunities before designing automation. NOT FOR: choosing and building the target architecture (suede-ops-architecture), auditing requester-supplied figures, or unapproved system access.

FileKindSize
CARD.mdmarkdown4.6 KB
SKILL.mdskill md12.3 KB
agents/openai.yamlconfig719 B
evals/evals.jsonconfig4.3 KB
references/interview-guide.mdmarkdown4.6 KB
references/worked-contract.mdmarkdown2.4 KB

references/interview-guide.md

View on GitHub
# Interview Guide

The question bank for Step 1. Run a walkthrough, not a survey: ask what happens
next until the process ends, then come back for the questions below that the
walkthrough did not already answer.

- [Ground rules](#ground-rules)
- [The core bank](#the-core-bank)
- [Role variants](#role-variants)
- [Turning answers into a floor map](#turning-answers-into-a-floor-map)
- [The two-day follow-up](#the-two-day-follow-up)
- [Answers that need a second pass](#answers-that-need-a-second-pass)

## Ground rules

**Interview the performer.** The manager describes the process as designed; the
person doing it describes the process as run. Both go in the record, and the
performer's version is the floor map.

**Ask about the last twenty cases, not about typical behavior.** "How often does
that happen?" returns an adjective. "Of the last twenty, how many took that
path?" returns a number.

**Never sell the future system during the interview.** Someone who believes a
tool is being justified describes the process they think you want. Someone who
believes the process is being recorded describes it.

**Say what the record is for and who reads it.** Costs get attached to processes,
not to people, and saying so up front is what makes the workaround questions
answerable.

**Record only with agreement in the session.** Written notes when anyone
declines.

## The core bank

Seven questions carry most assessments.

1. Walk me through this from the very beginning to the very end, including the
   steps that feel too small to mention.
2. Where does the information come from when you start, and where does it go
   when you finish?
3. What do you type or copy by hand more than once?
4. Where does work sit and wait, and who is it waiting on?
5. What do you double-check before you trust a number in a system?
6. What have you built for yourself that nobody officially knows about?
7. If volume doubled next quarter, what would break first?

Question 6 needs the ground rules to have landed. Question 7 returns the most
per minute of any question here.

## Role variants

Add to the core bank by what the person actually touches.

| Role | Add |
|---|---|
| Intake and coordination | What arrives incomplete, and what do you do while waiting for the rest? |
| Delivery and fulfillment | What do you rebuild each time because the last version cannot be found? |
| Finance and billing | What do you reconcile by hand, and between which two systems? |
| Support | Which answers do you retype, and where do you keep your own copy of them? |
| Management | Which question from leadership takes longest to answer, and what do you assemble to answer it? |
| Leadership | Which number do you not fully trust, and what do you do instead of trusting it? |

The management and leadership rows produce the reporting-lag findings. They do
not produce the floor map.

## Turning answers into a floor map

Each interview yields four record types. Keep them separate: they rank
differently in Step 4.

- **Steps**: what happens, in order, with who does it and what system it touches.
- **Waits**: where work sits, what it waits on, and for how long.
- **Re-entry**: the same information typed into more than one place.
- **Workarounds**: the unofficial spreadsheet, the private checklist, the
  personal folder. These name the silos that no software inventory will find.

A step that maps to no system is tribal knowledge. Record the holder, because
that is a key-person risk, and carry it forward rather than treating it as an
inventory miss.

## The two-day follow-up

Send the same three questions two days after every interview. The recall is
reliably better than anything the live session produced.

1. What did you remember after we spoke?
2. Which step happens rarely enough that it did not come up?
3. What did you describe as the normal path that has an exception you skipped?

The quarterly and annual exceptions surface here almost every time, and they are
the ones a system designed from the live session alone will fail on in month
four.

## Answers that need a second pass

Four responses that mean the question has not landed yet.

**"It depends."** A rule exists and has not been written down. Ask for the last
three cases and what decided each one; usually a lookup table falls out.

**"It just works."** The person has stopped seeing a workaround they perform
daily. Ask them to do it while you watch instead of describing it.

**"You'd have to ask <name>."** Tribal knowledge with a named holder. Record it
as key-person risk immediately, then go ask that person.

**"We're supposed to..."** The gap between the leadership map and the floor map,
said out loud. Ask what happens instead, and record both versions.
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall