everyinc__compound-engineering-plugin

skills/everyinc__compound-engineering-plugin

mirror of https://github.com/EveryInc/compound-engineering-plugin.git synced 2026-09-19 01:09:55 +08:00

Files

T

History

Trevin Chow 924cfea3a2 fix(ce-compound): auto-select mode and probe session history

Interactive ce-compound asked the user two questions they can't
usefully answer: Full vs Lightweight, and whether to search session
history. Mode depends on remaining context budget the agent can
observe but the user can't; session-history value depends on unrelated
prior sessions neither party knows about a priori. Both are now
resolved by the agent instead of a prompt.

- Full vs Lightweight is auto-selected (default Full; Lightweight only
  under real context pressure) and the choice is reported on the first
  output line.
- Session history runs as an automatic two-stage probe in Full mode:
  the cheap discovery+metadata pass always runs in parallel, escalating
  to extraction+synthesis only when a candidate clears a relevance bar
  (branch match or >=2 keyword hits).
- Dropped the "What's next?" menu; cross-doc maintenance stays routed to
  ce-compound-refresh via the refresh recommendation line.

The only surviving interactive prompt is the discoverability consent,
which edits a tracked instruction file. Headless mode unchanged.

2026-07-11 01:26:52 -07:00

brainstorms

refactor(plugin): make plugin skills-only and root-native (#967 )

2026-06-22 00:19:45 -07:00

ideation

feat(ce-pov): add project-grounded verdict skill (#1020 )

2026-06-29 15:49:52 -07:00

plans

feat(ce-plan,ce-brainstorm): dispatch reasoning-heavy steps to Fable on Claude Code (#1091 )

2026-07-09 17:59:54 -07:00

residual-review-findings

feat(ce-explain): add personal learning explainer skill with check-in loop (#1058 )