From praxis
Use when the goal is vague or vision-level and can't be defined well enough for design questions.
How this skill is triggered — by the user, by Claude, or both
Slash command
/praxis:discoverThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
`<gate>` No design, no code, no spec until "confirmed direction" is filled and user confirms. `</gate>`
<gate> No design, no code, no spec until "confirmed direction" is filled and user confirms. </gate>
The goal of discover is to turn a vague goal into a problem statement concrete enough for design. Ask one question per turn.
docs/discovery/YYYY-MM-DD-<topic>.md## Goal
<original vague goal>
## Hypotheses
- [ ] <hypothesis A>
- [x] <hypothesis B> → validated: <evidence>
- [~] <hypothesis C> → invalidated: <reason>
## Investigation plan
Contingent steps — update as findings arrive. Not a build plan.
- [x] Phase 1: <what> → <what to decide after>
- [ ] Phase 2: <what> (scope determined by Phase 1)
## Experiments
### YYYY-MM-DD — <experiment name>
what: <what was done>
saw: <what was observed>
conclusion: <what this means>
## Confirmed direction
<fill one of:>
- problem statement: who / what problem / what context / success signal
- knowledge spec: data structures, protocols, or system behavior confirmed
## Ruled out
- <idea> — <why ruled out>
Discovery notes accumulate across sessions. Check existing notes before starting.
spikes/YYYY-MM-DD-<topic>/Spike code validates a hypothesis, not a feature.
README.md stating which hypothesis it tests.design → plan → tdd — no direct promotion.<gate>
docs/discovery/YYYY-MM-DD-<topic>.md must exist with "Confirmed direction" filled.
Then choose the exit with the user:
design, referencing the discovery note.docs/discovery/<topic>-spec.md directly. No design, no plan, no archive (that skill is for staging specs only).
</gate>If the user decides not to proceed:
## Abandoned — YYYY-MM-DD
reason: <why>
validated: <what was confirmed>
invalidated: <what was ruled out>
spikes/<topic>/ — learnings are in the note.docs/discovery/ permanently as a record.npx claudepluginhub ouonet/praxis --plugin praxisProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.