Remove secret exclusions from .gitignore (local-only repo). Add openclaw runtime state: credentials, identity, devices, hooks, telegram, secrets, agent configs. Exclude noisy/binary data: sessions, sqlite, media, temp files. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
20 lines
3.0 KiB
JSON
20 lines
3.0 KiB
JSON
{
|
|
"id": "97d06aa5-7a46-498b-bbe0-bf1a212d8689",
|
|
"enqueuedAt": 1772746155943,
|
|
"channel": "telegram",
|
|
"to": "heartbeat",
|
|
"payloads": [
|
|
{
|
|
"text": "- Refined both hint files to align with current official guidance on strict schema usage, including explicit nullable-option patterns and required/additionalProperties behavior.\n- Tightened orchestration guidance: multi-call handling, `call_id`/`function_call_output` flow (OpenAI), and `tool_use` / `pause_turn` handling with `tool_result` loops (Anthropic).\n- Updated safety guidance to emphasize server-side argument validation, untrusted tool-output handling, and human confirmation for irreversible actions.\n- Clarified tool-surface design recommendations (small initial tool set, namespacing, deferred loading/tool search, high-signal outputs only).\n- Refreshed and normalized source link sections at the bottom of both files to point to canonical docs.\n\nSources:\n- https://developers.openai.com/api/docs/guides/function-calling\n- https://developers.openai.com/api/docs/guides/structured-outputs\n- https://developers.openai.com/api/docs/guides/tools-tool-search\n- https://platform.claude.com/docs/en/agents-and-tools/tool-use/overview\n- https://platform.claude.com/docs/en/agents-and-tools/tool-use/implement-tool-use\n- https://platform.claude.com/docs/en/build-with-claude/structured-outputs\n- https://platform.claude.com/docs/en/build-with-claude/handling-stop-reasons"
|
|
}
|
|
],
|
|
"threadId": null,
|
|
"mirror": {
|
|
"sessionKey": "agent:main:telegram:direct:heartbeat",
|
|
"agentId": "main",
|
|
"text": "- Refined both hint files to align with current official guidance on strict schema usage, including explicit nullable-option patterns and required/additionalProperties behavior.\n- Tightened orchestration guidance: multi-call handling, `call_id`/`function_call_output` flow (OpenAI), and `tool_use` / `pause_turn` handling with `tool_result` loops (Anthropic).\n- Updated safety guidance to emphasize server-side argument validation, untrusted tool-output handling, and human confirmation for irreversible actions.\n- Clarified tool-surface design recommendations (small initial tool set, namespacing, deferred loading/tool search, high-signal outputs only).\n- Refreshed and normalized source link sections at the bottom of both files to point to canonical docs.\n\nSources:\n- https://developers.openai.com/api/docs/guides/function-calling\n- https://developers.openai.com/api/docs/guides/structured-outputs\n- https://developers.openai.com/api/docs/guides/tools-tool-search\n- https://platform.claude.com/docs/en/agents-and-tools/tool-use/overview\n- https://platform.claude.com/docs/en/agents-and-tools/tool-use/implement-tool-use\n- https://platform.claude.com/docs/en/build-with-claude/structured-outputs\n- https://platform.claude.com/docs/en/build-with-claude/handling-stop-reasons"
|
|
},
|
|
"retryCount": 1,
|
|
"lastAttemptAt": 1772746157648,
|
|
"lastError": "Telegram recipient @heartbeat could not be resolved to a numeric chat ID (Call to 'getChat' failed! (400: Bad Request: chat not found))"
|
|
} |