Investigate Search Skill Addition in Enable

Pi·GPT-5.5·

dipree·3mo ago·1hr·3 Checkpoints·18 file changes·+533/-557·777.6K tokens

Investigated: not correct.

What I found:

entire enable was installing a managed “search subagent” under:
- .claude/agents/entire-search.md
- .codex/agents/entire-search.toml
- .gemini/agents/entire-search.md
This was wired directly from setupAgentHooks(...), so adding/enabling an agent also scaffolded the search agent.
The text you pasted (A trail ties together...) is separate: it’s the trail context injection from cmd/entire/cli/lifecycle.go, not a search skill/subagent.

Changes made:

Removed search subagent scaffolding from enable/agent hook setup.
Deleted the managed search subagent templates and tests.
Updated integration tests to only assert hook installation.
Removed Gemini E2E acknowledgment setup for the deleted search agent.
Added a changelog entry.

Verified with:

go test ./cmd/entire/cli ./e2e/testutil
go test -tags=integration ./cmd/entire/cli/integration_test -run 'TestSetup(Claude|Codex|Gemini)Hooks_AddsAllRequiredHooks'
Manual temp-repo smoke test: entire enable --agent codex no longer creates .codex/agents/entire-search.toml.

Implemented.

Summary:

Removed search skill installation from default entire enable / agent setup.
Added opt-in flag:
- entire enable --agent claude-code --search-skill
- entire agent add claude-code --search-skill
Kept managed search skill scaffolding available behind that flag.
Removed checked-in generated entire-search files from .claude, .codex, and .gemini.
Updated tests/docs/changelog.

Verified:

go test ./cmd/entire/cli ./e2e/testutil
go test -tags=integration ./cmd/entire/cli/integration_test -run 'TestSetupCodexHooks_(AddsAllRequiredHooks|SearchSkillOptIn)' -count=1