Skip to content

Conversation

@bokelley
Copy link
Contributor

@bokelley bokelley commented Jan 3, 2026

Summary

  • Add GitHub issue offer for open-source agent test failures - when users test agents via Addie and the tests fail on known open-source agents, Addie offers to help file a GitHub issue to the appropriate repository
  • Refactor to use @adcp/client/testing library directly instead of local wrapper
  • Add call_adcp_tool handler for raw AdCP task execution

Known Open-Source Agents

The feature recognizes these agents:

  • test-agent.adcontextprotocol.org → adcontextprotocol/salesagent
  • wonderstruck.sales-agent.scope3.com → adcontextprotocol/salesagent
  • creative.adcontextprotocol.org → adcontextprotocol/creative-agent

Test plan

  • TypeScript compiles without errors
  • All existing tests pass
  • Manually tested via Addie chat - ran discovery scenario against test agent
  • Verified GitHub issue offer message appears when tests fail
  • Verified draft_github_issue tool creates pre-filled GitHub issue link

🤖 Generated with Claude Code

bokelley and others added 9 commits January 2, 2026 14:44
- Add agent_contexts and agent_test_history tables for persisting agent URLs
- Implement AES-256-GCM encrypted token storage with org-specific keys
- Add save_agent, list_saved_agents, remove_saved_agent tools to Addie
- Extend test_adcp_agent to auto-lookup saved credentials and record history
- Add agent-tester.ts with E2E test scenarios including response_consistency

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>
PORT (internal server port, 8080 in Docker) should take precedence over
CONDUCTOR_PORT (external mapping) when making internal API calls.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>
Document the new E2E agent testing capabilities:
- Testing scenarios (discovery, full_sales_flow, etc.)
- Edge case scenarios (error_handling, validation)
- Behavioral analysis features
- Sales agent compliance checklist

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>
When users test agents via Addie and the tests fail on a known open-source
agent (test-agent.adcontextprotocol.org, wonderstruck.sales-agent.scope3.com,
or creative.adcontextprotocol.org), Addie now offers to help file a GitHub
issue to the appropriate repository.

Also refactors to use @adcp/client/testing library directly instead of local
agent-tester.ts wrapper:
- Imports runAgentTests, formatTestResults, createTestClient from library
- Adds call_adcp_tool handler for raw task execution
- Removes redundant local agent-tester.ts (2570 lines)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>
- Update homepage "Test AdCP" button to route to Addie chat (/chat.html)
  instead of external testing.adcontextprotocol.org
- Add image support to chat.html with clickable images that open in new tab
- Add iframe support for creative previews with sandbox and styling
- Add CSS styles for embedded images and creative preview containers

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>
- Add "Try AdCP with a test agent" as suggested prompt in Addie chat
  (replaces "Become a member" to encourage testing)
- Rename 079_agent_contexts.sql to 109_agent_contexts.sql to resolve
  migration number conflict with main branch

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>
- Support ?prompt= query parameter to pre-fill and auto-send a message
- Update homepage "Test with Addie" button to include prompt parameter
- Waits for Addie to be ready before sending, with 10s timeout
- Cleans URL after sending to avoid re-triggering on refresh

Now clicking "Test with Addie" on homepage will automatically trigger
the test agent flow: /chat.html?prompt=Try%20AdCP%20with%20a%20test%20agent

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>
The script file is check-testable-snippets.cjs but the workflow
was referencing check-testable-snippets.js, causing CI failures.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>
@bokelley bokelley merged commit 9a3a6fb into main Jan 3, 2026
7 checks passed
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.

2 participants