Skip to content

Onboarding + discoverability: interactive lens picker, full toolbox, exhaustive mining#19

Merged
0xadvait merged 1 commit into
mainfrom
onboarding-discoverability
Jun 6, 2026
Merged

Onboarding + discoverability: interactive lens picker, full toolbox, exhaustive mining#19
0xadvait merged 1 commit into
mainfrom
onboarding-discoverability

Conversation

@0xadvait

@0xadvait 0xadvait commented Jun 6, 2026

Copy link
Copy Markdown
Owner

Users couldn't tell how much is in doxa. This makes capabilities discoverable on both surfaces (terminal + the agent), and fixes the miner under-extracting.

Terminal

  • doxa init now asks the user to pick a lens from the library (numbered picker, like the provider prompt) instead of silently defaulting. --lens-template <name> still skips straight in.
  • The bare doxa landing teases the breadth and points to doxa guide.
  • doxa guide gains "THE FULL TOOLBOX" -- every command grouped by phase (lenses / ingest / ask / keep-honest).

Agent (the "in the agent itself" part)

  • SKILL.md rewritten into a comprehensive, proactive playbook with a capabilities table -- an installed agent now knows about lenses, the --via fetchers + --mode/--prompt/--yolo, eval, doctor, sources, and semantic search, not just query+ingest.
  • AGENTS.md gains a capabilities map + the lens library + doctor + --yolo.

Mining (no assumed counts)

  • The extraction prompt no longer implies a belief/quote count. It mines every distinct belief the lens surfaces per chunk -- no target number, no cap, no pad, scaled to the source -- so a lens template actually surfaces everything it asks for.
  • tests for the picker, the discoverability surfaces, and the exhaustive-mining prompt. Full suite 109 passing.

🤖 Generated with Claude Code

…exhaustive mining

Users couldn't tell how much doxa can do. Make it discoverable on both surfaces,
and stop the miner from under-extracting.

Terminal:
- `doxa init` now presents the lens library and asks the user to pick one (numbered
  picker, like the provider prompt) instead of defaulting silently -- "don't invent
  your first lens." `--lens-template` still skips straight to a template.
- The bare `doxa` landing teases the breadth (lenses, --via fetchers, --answer/--json/
  hybrid, eval/doctor) and points to `doxa guide`.
- `doxa guide` gains a "THE FULL TOOLBOX" section grouping every command by phase
  (lenses / ingest / ask / keep-honest).

Agent:
- SKILL.md rewritten into a comprehensive, proactive playbook with a capabilities
  table -- the installed agent now knows about lenses, fetchers + --mode/--prompt/--yolo,
  eval, doctor, sources, and semantic search, not just query+ingest.
- AGENTS.md gains a capabilities-at-a-glance map, the lens library, doctor, and --yolo.

Mining:
- The extraction prompt no longer implies a belief/quote count. It now mines EVERY
  distinct belief the lens surfaces in each chunk -- no target number, no cap, no pad,
  scaled to the source -- so a template actually surfaces everything it asks for.

+ picker/discoverability/exhaustive-mining tests. Full suite 109 passing.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
@0xadvait
0xadvait merged commit e22fb13 into main Jun 6, 2026
@0xadvait
0xadvait deleted the onboarding-discoverability branch June 6, 2026 21:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant