discovery-define-outcomeslisted
Install: claude install-skill timothymeyers/discovery-tools
# discovery-define-outcomes
## Modes
- **Proposal-only** (called by `discovery-project-setup`): return ops in
that skill's proposal format and stop. Do not mutate.
- **Standalone**: gather, propose, approve, apply using
`../discovery-project-setup/references/{gather-state,proposal-format,apply,git-ledger,retirement}.md`.
Ask questions as plain chat, not the question modal.
## Inputs
Purpose (`purpose_read get`), current outcomes (`outcome.list`), staged or
ingested documents, and the user's stated goals.
## Drafting rules
- **Count:** default 4–5. Go up to 10 only when the material clearly
supports distinct, separately verifiable results. Exclude `[Superseded]`
outcomes from the count. Over 10 means split the project or merge outcomes;
say which.
- **Type**, one of: `Artifact` (a thing produced), `Measurement` (a
quantified result), `Decision` (a choice made with evidence), `Knowledge`
(a finding or answer), `Capability` (something that now works).
- **Each outcome has:**
- a title: a short result statement, not an activity ("Validated N-1
screen on synthetic grid", not "Run N-1 screen")
- evidence: what artifact or measurement shows it's done, with a threshold
when one exists (use LaTeX for quantities, e.g. $\le 5\%$)
- a source citation: the document section or the user's words
- a rubric draft: 2–4 questions, `yesNo` or `scale`, and passing criteria
- **Quality checks**, applied before proposing. Flag outcomes that fail them:
- Verifiab