Update agent-integration skill docs

[alishakawaguchi]

Summary

• Improved agent-integration skill documentation with clearer parameter descriptions and derivation guidance
• Rewrote implementer docs with E2E-first TDD approach (unit tests written last) and detailed implementation patterns
• Simplified researcher and test-writer docs for clarity and consistency with actual codebase patterns

Test plan

• Verify /agent-integration skill triggers correctly with updated SKILL.md
• Run through researcher phase to confirm updated instructions are clear
• Run through test-writer phase to confirm updated instructions work
• Run through implementer phase to confirm E2E-first TDD flow works

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Documentation-only changes, but they materially alter the recommended agent integration workflow (E2E-first tiers, unit tests last) and required artifacts (AGENT.md), so incorrect guidance could slow or misdirect future integrations.

Overview
Updates the agent-integration skill docs to standardize on E2E-first TDD: build and iterate by running existing ForEachAgent E2E tests in explicit tiers, defer all unit tests until the end, and use /debug-e2e + frequent /commit checkpoints.

Shifts Phase 1 output from a generic compatibility report to a required agent-specific cmd/entire/cli/agent/$AGENT_PACKAGE/AGENT.md one-pager that becomes the single source of truth for later phases, updates parameter conventions (AGENT_PACKAGE, AGENT_KEY), and narrows scope to hook-capable agents.

Refocuses Phase 2 (write-tests) to only create/verify the E2E agent runner (no new scenarios), and rewrites researcher guidance to rely on agent-guide.md plus external probing while explicitly avoiding other internal source reads.

^{Written by Cursor Bugbot for commit ee048ec. Configure here.}

[Copilot]

Pull request overview

Updates the agent-integration skill documentation to standardize around an E2E-first TDD workflow, with a single-source-of-truth agent one-pager (AGENT.md) feeding the runner and implementation phases.

Changes:

• Refocused the test-writer phase on creating only an E2E agent runner (no new scenarios / no unit tests).
• Reworked the researcher phase to output a structured AGENT.md one-pager for downstream phases.
• Rewrote implementer guidance to enforce E2E-first development with unit tests written last.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 10 comments.

File	Description
.claude/skills/agent-integration/test-writer.md	Shifts phase to “runner only” and adds compile/registration verification steps.
.claude/skills/agent-integration/researcher.md	Replaces feasibility report output with a persisted AGENT.md one-pager template.
.claude/skills/agent-integration/implementer.md	Introduces strict E2E-tier workflow and consolidates unit tests into the final step.
.claude/skills/agent-integration/SKILL.md	Updates skill description, parameter derivation guidance, and E2E-first TDD rule set.