Which skill carries which step, what each decides for a person, and what each refuses to do
The measurement loop in “A Discoverability Score Measures Whether Finders Who Don’t Know Your Name Meet You” has seven steps and nine prompts. They are packaged as three skills for an AI harness, grouped by the outcome a practitioner wants, not one per step. The full prompts live inside the skills and nowhere else, so each has one source. The toolkit download includes the three skill files under skills/.
A skill does the repeatable work and stops where a person must decide. Each one asks for the person’s decision, records it, and ends with a recommendation and its reasoning. Where the evidence is too thin for any recommendation, it says so and names what would resolve it. None of them computes a score: the toolkit’s scripts do.
| Skill | The outcome | Prompts it carries | What it leaves to a person |
|---|---|---|---|
|
A frozen benchmark drawn from the Finder Specification |
Build the Finder Specification; generate the finder questions; review the corpus as an adversary |
Who the finder is, which questions matter, and the decision to freeze |
|
A scorecard, an executive page, and a practitioner report for a baseline or a retest |
Analyze the answers; write the Executive Finding |
Checking the answers against the raw text, rating the evidence, and capturing answers in a way the engine’s maker permits |
|
The alternatives recorded, the gaps diagnosed, the interventions chosen, and the retest prepared |
Analyze the alternatives that appear; diagnose the gaps; design the minimum interventions; prepare the retest |
Which findings to act on, which interventions are ethical, and what the result means for the business |
What the three share
-
Raw answers are never edited. The skills store each answer whole before they read it.
-
The
archon scorecommands are the only source of a number. A skill that is asked for a score runsarchon score showand reports what it says, including "not computed." -
Every finding has three labels. Observation, hypothesis, and experiment, with the observation that would disprove the hypothesis.
-
No cause is stated as fact. A mention is not a recommendation, and a rise in the score is not success until someone has asked whether the right finders now meet the subject, and why.
-
No intervention manufactures evidence. A change that is not tied to an observed gap is not proposed.
Release
Per the press’s rule, a skill does not reach a reader before the book’s own publish decision is recorded. The skills are in the press’s shared skills library and are linked to this book. They ship with the book’s decision, not ahead of it.