Hermes Agent backup - 2026-07-21 09:44:07

logs/agent.log

@@ -2057,3 +2057,16 @@
 2026-07-21 09:42:36,433 INFO [20260721_091814_2fcbda] agent.conversation_loop: API call #15: model=deepseek-ai/deepseek-v4-pro provider=nvidia in=177983 out=216 total=178199 latency=47.6s cache=24576/177983 (14%)
 2026-07-21 09:42:36,433 INFO [20260721_091814_2fcbda] agent.tool_executor: tool todo completed (0.00s, 561 chars)
 2026-07-21 09:42:36,461 INFO run_agent: OpenAI client created (chat_completion_stream_request, shared=False) thread=Thread-23 (_call):139766129485504 provider=nvidia base_url=https://integrate.api.nvidia.com/v1 model=deepseek-ai/deepseek-v4-pro
+2026-07-21 09:43:44,916 INFO run_agent: OpenAI client closed (stream_request_complete, shared=False, tcp_force_closed=0) thread=Thread-23 (_call):139766129485504 provider=nvidia base_url=https://integrate.api.nvidia.com/v1 model=deepseek-ai/deepseek-v4-pro
+2026-07-21 09:43:44,916 INFO [20260721_091814_2fcbda] agent.conversation_loop: API call #16: model=deepseek-ai/deepseek-v4-pro provider=nvidia in=178373 out=838 total=179211 latency=68.5s cache=160000/178373 (90%)
+2026-07-21 09:43:44,925 INFO [20260721_091814_2fcbda] agent.tool_executor: tool skill_manage completed (0.01s, 327 chars)
+2026-07-21 09:43:44,953 INFO run_agent: OpenAI client created (chat_completion_stream_request, shared=False) thread=Thread-24 (_call):139766129485504 provider=nvidia base_url=https://integrate.api.nvidia.com/v1 model=deepseek-ai/deepseek-v4-pro
+2026-07-21 09:44:03,640 INFO run_agent: OpenAI client closed (stream_request_complete, shared=False, tcp_force_closed=0) thread=Thread-24 (_call):139766129485504 provider=nvidia base_url=https://integrate.api.nvidia.com/v1 model=deepseek-ai/deepseek-v4-pro
+2026-07-21 09:44:03,641 INFO [20260721_091814_2fcbda] agent.conversation_loop: API call #17: model=deepseek-ai/deepseek-v4-pro provider=nvidia in=179314 out=92 total=179406 latency=18.7s
+2026-07-21 09:44:03,642 INFO [20260721_091814_2fcbda] tools.terminal_tool: Creating new local environment for task default...
+2026-07-21 09:44:03,645 WARNING [20260721_091814_2fcbda] tools.env_passthrough: env passthrough: refusing to register Hermes provider credential 'GEMINI_API_KEY' from config.yaml (blocked by _HERMES_PROVIDER_ENV_BLOCKLIST). Operator configuration must not override the execute_code sandbox's credential scrubbing; see GHSA-rhgp-j443-p4rf.
+2026-07-21 09:44:03,645 WARNING [20260721_091814_2fcbda] tools.env_passthrough: env passthrough: refusing to register Hermes provider credential 'GOOGLE_API_KEY' from config.yaml (blocked by _HERMES_PROVIDER_ENV_BLOCKLIST). Operator configuration must not override the execute_code sandbox's credential scrubbing; see GHSA-rhgp-j443-p4rf.
+2026-07-21 09:44:03,698 INFO [20260721_091814_2fcbda] tools.environments.base: Session snapshot created (session=1c45b291f136, cwd=/home/appuser)
+2026-07-21 09:44:03,698 INFO [20260721_091814_2fcbda] tools.terminal_tool: local environment ready for task default
+2026-07-21 09:44:03,871 INFO [20260721_091814_2fcbda] agent.tool_executor: tool terminal completed (0.23s, 45 chars)
+2026-07-21 09:44:03,901 INFO run_agent: OpenAI client created (chat_completion_stream_request, shared=False) thread=Thread-28 (_call):139766108505792 provider=nvidia base_url=https://integrate.api.nvidia.com/v1 model=deepseek-ai/deepseek-v4-pro

logs/errors.log

@@ -222,3 +222,5 @@
 
 [Tool loop warning: same_tool_failure_warning; count=3; skill_view has failed 3 tim
 2026-07-21 09:41:48,793 WARNING [20260721_091814_2fcbda] agent.tool_executor: Tool skill_view returned error (0.27s): {"success": false, "error": "Skill 'serving-llms-vllm' not found.", "available_skills": ["apikey-image-gen", "dogfood", "grok-image-to-video", "hyperframes", "markdown-viewer", "remotion", "yuanbao", 
+2026-07-21 09:44:03,645 WARNING [20260721_091814_2fcbda] tools.env_passthrough: env passthrough: refusing to register Hermes provider credential 'GEMINI_API_KEY' from config.yaml (blocked by _HERMES_PROVIDER_ENV_BLOCKLIST). Operator configuration must not override the execute_code sandbox's credential scrubbing; see GHSA-rhgp-j443-p4rf.
+2026-07-21 09:44:03,645 WARNING [20260721_091814_2fcbda] tools.env_passthrough: env passthrough: refusing to register Hermes provider credential 'GOOGLE_API_KEY' from config.yaml (blocked by _HERMES_PROVIDER_ENV_BLOCKLIST). Operator configuration must not override the execute_code sandbox's credential scrubbing; see GHSA-rhgp-j443-p4rf.

metadata.json

@@ -1,5 +1,5 @@
 {
-  "timestamp": "2026-07-21T09:42:51.436567",
+  "timestamp": "2026-07-21T09:44:07.872115",
   "version": "0.10.0",
   "hermes_home": "/data/.hermes"
 }

skills/github/github/SKILL.md

@@ -0,0 +1,63 @@
+---
+name: github
+description: Complete GitHub operations — auth, repos, PRs, issues, code review, CI, all via gh CLI or REST fallback.
+version: 1.0.0
+author: Hermes Agent
+license: MIT
+platforms: [linux, macos, windows]
+metadata:
+  hermes:
+    tags: [GitHub, Git, gh-cli, Pull-Requests, Issues, Code-Review, CI, Repositories, Authentication]
+---
+
+# GitHub — Full Operations
+
+Everything for working with GitHub from Hermes: authentication, repository management, pull request lifecycle, issue tracking, code review, and CI monitoring. Every section supports two paths: **`gh` CLI** (preferred when installed) and **pure `git` + `curl`** (always available, no installation needed).
+
+## Auth Setup
+
+See `references/auth.md` for the full authentication guide: HTTPS tokens, SSH keys, `gh auth login`, and the auth-detection helper that all GitHub workflows should start with.
+
+## Repository Management
+
+See `references/repo-management.md` for cloning, creating, forking, configuring repos, releases, and GitHub Actions management.
+
+## Pull Request Workflow
+
+See `references/pr-workflow.md` for the complete PR lifecycle: branch creation, conventional commits, PR creation, CI monitoring, auto-fixing failures, and merging.
+
+## Issues Management
+
+See `references/issues.md` for creating, searching, triaging, labeling, assigning, commenting, closing/reopening, and bulk-operating on issues.
+
+## Code Review
+
+See `references/code-review.md` for reviewing local changes (pre-push), reviewing PRs on GitHub, leaving inline comments, submitting formal reviews, and the review checklist.
+
+## Common Setup Block
+
+Every GitHub workflow should start with the auth-detection block from `references/auth.md`. It determines whether to use `gh` or `git` + `curl` for the session.
+
+## Verification Checklist
+
+- [ ] Auth is configured (gh auth status or GITHUB_TOKEN available)
+- [ ] Inside a git repository with a GitHub remote (for most operations)
+- [ ] Owner/repo extracted from remote URL
+- [ ] Correct path (gh vs git+curl) selected based on availability
+
+## Common Pitfalls
+
+1. **GitHub disabled password auth** — always use a personal access token or SSH, never account password
+2. **`gh: command not found`** — fall back to `git` + `curl` with a `GITHUB_TOKEN`
+3. **Cached stale credentials** — run `git credential reject` then re-authenticate
+4. **PR creation without `--head` branch** — ensure the branch was pushed with `git push -u origin HEAD` first
+5. **Merge conflict markers in diffs** — always scan diffs for `<<<<<<` / `>>>>>>` patterns
+6. **Forgetting to extract owner/repo** — many curl commands need `$OWNER/$REPO`; always run the extract block
+
+## References
+
+- `references/auth.md` — Authentication setup (HTTPS tokens, SSH keys, gh login, auth detection)
+- `references/repo-management.md` — Repository management (clone, create, fork, settings, releases, CI)
+- `references/pr-workflow.md` — Pull request lifecycle
+- `references/issues.md` — Issues management
+- `references/code-review.md` — Code review (local + PR)