Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/bug-ops/zeph/setup-guidenpx skills add bug-ops/zeph --skill setup-guidegit clone --depth 1 https://github.com/bug-ops/zephWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00084 | $0.02658 |
| Opus 5 | $0.00042 | $0.01329 |
| Sonnet 5 | $0.00017 | $0.00532 |
| Haiku 4.5 | $0.00008 | $0.00266 |
Grade A, and why
setup-guide scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
export ZEPH_TOOLS_SHELL_ALLOWED_COMMANDS=curl,wget How it starts
The opening of the file, as written. The whole thing — 296 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Setup Guide
LLM Provider
Ollama (default):
export ZEPH_LLM_PROVIDER=ollama
export ZEPH_LLM_BASE_URL=http://localhost:11434
export ZEPH_LLM_MODEL=mistral:7b
Claude:
export ZEPH_LLM_PROVIDER=claude
export ZEPH_CLAUDE_API_KEY=sk-ant-...
Cloud model settings in config/default.toml:
llm.cloud.model(default:claude-sonnet-4-5-20250929)llm.cloud.max_tokens(default: 4096)
OpenAI (or any OpenAI-compatible API):
export ZEPH_LLM_PROVIDER=openai
export ZEPH_OPENAI_API_KEY=sk-...
Config in config/default.toml:
[llm.openai]
base_url = "https://api.openai.com/v1"
model = "gpt-5.2"
max_tokens = 4096
embedding_model = "text-embedding-3-small"
reasoning_effort = "medium" # low, medium, high (for reasoning models)
llm.openai.base_url: API endpoint (change for Together, Groq, Fireworks, etc.)llm.openai.model: chat model namellm.openai.max_tokens: max response tokens (default: 4096)llm.openai.embedding_model: optional, enables embeddings supportllm.openai.reasoning_effort: optional,low/medium/highfor reasoning models (o3, etc.)
Embeddings
export ZEPH_LLM_EMBEDDING_MODEL=qwen3-embedding
Used for skill matching and semantic memory. Pull model first:
ollama pull qwen3-embedding
Memory
SQLite storage:
export ZEPH_SQLITE_PATH=.zeph/data/zeph.db
Config: memory.history_limit (default: 50) — recent messages loaded into context.
Semantic Memory (Qdrant)
export ZEPH_MEMORY_SEMANTIC_ENABLED=true
export ZEPH_QDRANT_URL=http://localhost:6334
export ZEPH_MEMORY_RECALL_LIMIT=5
Start Qdrant:
docker compose up -d qdrant
When semantic memory is enabled and Qdrant is reachable, skill embeddings are persisted in a zeph_skills collection. On startup, only changed skills are re-embedded (BLAKE3 content hash comparison). The Qdrant HNSW index is used for skill matching instead of in-memory cosine similarity. If Qdrant is unavailable, the agent falls back to in-memory matching.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 296 lines · 84 tokens per session scan A 65f906aadd96
setup-guide is a skill published in the GitHub repository bug-ops/zeph (57 stars, last pushed 7d ago), licensed MIT. It adds 84 tokens to every session and 2,658 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
verify
Run Chimeraforge's canonical verification gate end-to-end and report the real output before claiming work done or committing. Failing output gets pasted, fixed, and re-run — never summarized away.
free-model-manager
Free AI model management - discover, download, and manage free/open-source AI models from Ollama, HuggingFace, and other sources.
feishu
Work with Feishu or Lark bots, docs, sheets, bitables, approval flows, and OpenAPI/MCP setup without hardcoding credentials.
interview
Ask one useful structured question at a time only when material product/implementation choices are genuinely missing; remember answers and produce a brief/spec. Discoverable facts should be investigated instead of asked.
test
Detect the project’s test stack, run the narrowest useful tests, create tests when authorized, and report coverage/gaps honestly.
verify
Exercise the real app/API/CLI and collect observable evidence; tests alone do not count as end-to-end verification.