go-go-golems/inference-smoke-tests
.codex/skills/inference-smoke-tests/SKILL.md
Run repeatable inference smoke tests using geppetto/pinocchio example binaries (single-pass, streaming, tool-loop, OpenAI Responses thinking) including tmux-driven TUI tests. Use when refactors touch InferenceState/Session/EngineBuilder, tool calling loop, event sinks, provider request formatting, or when you need a quick 'does inference still work?' checklist.
$ install --global
skillsauthnpx skillsauth add go-go-golems/geppetto inference-smoke-testsInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 5, 2026, 1:58 PM9.9s3 files scanned
SKILL.md
- name:
- inference-smoke-tests
- description:
- Run repeatable inference smoke tests using geppetto/pinocchio example binaries (single-pass, streaming, tool-loop, OpenAI Responses thinking) including tmux-driven TUI tests. Use when refactors touch InferenceState/Session/EngineBuilder, tool calling loop, event sinks, provider request formatting, or when you need a quick 'does inference still work?' checklist.
Inference Smoke Tests
Quick Start (Recommended)
Run the fast suite (geppetto non-TUI + pinocchio agent TUI) via the bundled script:
bash geppetto/.codex/skills/inference-smoke-tests/scripts/run_smoke.sh --quick
If you need the full manual checklist, open:
geppetto/.codex/skills/inference-smoke-tests/references/playbook.md
Preconditions
- Ensure
OPENAI_API_KEYis set (for OpenAI Chat + OpenAI Responses). - Ensure Claude credentials are available (e.g.
ANTHROPIC_API_KEY) if you want the Claude tool-calling smoke step to pass. - Ensure
tmuxis installed (required for non-interactive TUI runs). - Expect costs: these tests make real API calls.
Workflow Decision Tree
- Validate provider “thinking” streaming (Responses)?
- Run
geppetto/cmd/examples/openai-toolsin--mode thinking.
- Validate tool loop orchestration?
- Run
geppetto/cmd/examples/generic-tool-calling.
- Validate Bubble Tea TUI event flow (thinking deltas + final)?
- Run
pinocchio/cmd/agents/simple-chat-agentin tmux.
- Validate Claude tool calling?
- Run
geppetto/cmd/examples/claude-toolswith--ai-api-type claude --ai-engine claude-haiku-4-5.
- Validate multi-turn chat state persistence?
- Run pinocchio TUI chat in tmux (manual) and/or pinocchio webchat in browser (manual).
What “Benefits From InferenceState” (Rules of Thumb)
Already benefits (multi-turn, cancel-sensitive, tool-loop, strict provider validation):
- pinocchio TUI chat (
pinocchio/cmd/pinocchio … --chat) - pinocchio agent TUI (
pinocchio/cmd/agents/simple-chat-agent …) - pinocchio webchat (
pinocchio/cmd/web-chat) - geppetto example runners that execute via
geppetto/pkg/inference/core.Session
Could benefit (optional; mainly consistency/cancel):
pinocchio/cmd/examples/simple-redis-streaming-inference(transport-focused; currentlyeng.RunInferencedirect)pinocchio/cmd/examples/simple-chat(exercises PinocchioCommand runner; could benefit indirectly if that runner standardizes onInferenceState)
Does not apply (not an inference runner):
geppetto/cmd/examples/citations-event-stream
Troubleshooting (Common Failure Modes)
“OpenAI Responses 400” errors
- Re-run with higher logging:
- Add
--log-level debug --with-callerwhere supported.
- Add
- Confirm you’re using the correct provider mode:
--ai-api-type openai-responses
- If the error mentions invalid parameter support (e.g.,
temperatureunsupported), it’s model-dependent; reduce parameters and retry.
TUI doesn’t submit the prompt
- Some TUIs submit on
Tab(notEnter). - Always capture logs to a file and confirm inference actually ran (look for
EventPartialCompletionStart,EventFinal).
References
When you need copy/paste commands for the full sweep, read:
geppetto/.codex/skills/inference-smoke-tests/references/playbook.md
When you need to find new example entry points, search:
rg -n "cmd/examples" -S geppetto/cmd/examples pinocchio/cmd/examples
rg -n "cmd/agents" -S pinocchio/cmd/agents
Related Skills
go-go-golems/ttmp/_templates
documentation
VerifiedTrustedCommunity
--- Title: {{TITLE}} Ticket: {{TICKET}} Status: draft Topics: {{TOPICS}} DocType: skill Intent: long-term Owners: {{OWNERS}} RelatedFiles: [] ExternalSources: [] Summary: > {{SUMMARY}} LastUpdated: {{DATE}} WhatFor: > <!-- 2–5 sentences: what outcome does this workflow enforce? --> WhenToUse: > <!-- Trigger conditions: "Use when ..." --> --- # {{TITLE}} ## Overview <!-- 2–5 sentences: why this skill matters and what it enforces --> ## When to Use <!-- concrete triggers + examples -->
84SKILL.mdUpdated Apr 5, 2026
go-go-golems/ttmp/_guidelines
development
VerifiedTrustedCommunity
# Guidelines: Skill Documents ## Purpose Skills are **disciplined workflows** written as documents (`DocType: skill`) that teach LLMs (and humans) *how to work*, not just what exists. A good skill turns “best practice” into a repeatable, enforceable process. Skills are meant to be discoverable via: - `docmgr skill list` (filter by topics, file/dir, ticket) - `docmgr skill show <query>` (load and apply a skill) ## Required Elements - **Frontmatter contract** - `DocType: skill` - `Title`: U
84SKILL.mdUpdated Apr 5, 2026
openclaw/taskflow
tools
VerifiedTrustedCommunity
Use when work should span one or more detached tasks but still behave like one job with a single owner context. TaskFlow is the durable flow substrate under authoring layers like Lobster, ACPX, plugins, or plain code. Keep conditional logic in the caller; use TaskFlow for flow identity, child-task linkage, waiting state, revision-checked mutations, and user-facing emergence.
357,764SKILL.mdUpdated Apr 10, 2026
openclaw/extensions/lobster
tools
VerifiedTrustedCommunity
# Lobster Lobster executes multi-step workflows with approval checkpoints. Use it when: - User wants a repeatable automation (triage, monitor, sync) - Actions need human approval before executing (send, post, delete) - Multiple tool calls should run as one deterministic operation ## When to use Lobster | User intent | Use Lobster? | | ------------------------------------------------------ | --------------------------
357,764SKILL.mdUpdated Apr 10, 2026