feat(composio-direct): add Composio direct integration plugin

[konard]

Summary

Implements the composio-direct plugin as specified in #105.

• 4 tools for direct Composio API access without MCP/SSE transport:

  • composio_search_tools — discover tools by query or toolkit name
  • composio_execute_tool — execute a single tool; returns structured auth_required error with connect_url on 401/403
  • composio_multi_execute — parallel batch execution with fail_fast and configurable concurrency
  • composio_auth_link — generate OAuth connect links (dynamic URL from Composio initiate endpoint with fallback)

• Retry logic — 3 attempts with exponential backoff (1 s, 2 s, 4 s) for network errors and 5xx responses

• Security — composio_api_key and OAuth tokens are never written to logs; action tools scoped to dm-only

• 30 automated tests (23 unit + 7 integration) covering success paths, auth errors, retries, batching, fail_fast, and validation

How to reproduce the issue and verify the fix

Prerequisite: Set COMPOSIO_API_KEY env var or configure composio_api_key secret in Teleton.

Manual verification checklist:

[ ] Plugin loads without errors
[ ] composio_search_tools query="github" returns a list of tools
[ ] composio_auth_link service="github" returns a valid connect URL
[ ] composio_execute_tool with unknown slug returns a helpful error
[ ] Auth errors (401) return { success: false, error: "auth_required", auth: { connect_url } }
[ ] composio_multi_execute with 2+ tools returns results in original order
[ ] Timeouts and retries: observable via debug logs

Test plan

• Run unit tests: node --test plugins/composio-direct/test/unit/composio-direct.test.js
• Run integration tests: node --test plugins/composio-direct/test/integration/composio-api.test.js
• All 30 tests pass (0 failures)
• Manual end-to-end test with a real COMPOSIO_API_KEY

🤖 Generated with Claude Code

Fixes #105

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

• Public pricing estimate: $1.307586
• Calculated by Anthropic: $1.307586 USD
• Difference: $0.000000 (+0.00%)

📊 Context and tokens usage:

Claude Sonnet 4.6:

• Max context window: 65.1K / 1M input tokens (7%)
• Max output tokens: 25.1K / 64K output tokens (39%)

Total input tokens: 54.8K + 1.8M cached
Total output tokens: 25.1K output
Cost: $1.127724

Claude Haiku 4.5:

• Context window: 320.6K / 200K tokens (160%)
• Max output tokens: 16.0K / 64K output tokens (25%)

Total input tokens: 60.1K + 260.4K cached
Total output tokens: 16.0K output
Cost: $0.179862

🤖 Models used:

• Tool: Anthropic Claude Code
• Requested: sonnet
• Main model: Claude Sonnet 4.6 (claude-sonnet-4-6)
• Additional models:
  • Claude Haiku 4.5 (claude-haiku-4-5-20251001)

📎 Log file uploaded as Gist (1034KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.