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 instructions/ozmasterai/torus-framework/claude-mdgit clone --depth 1 https://github.com/OZmasterAI/Torus-FrameworkWhat 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 | $0.01357 | $0.01357 |
| Opus 5 | $0.00678 | $0.00678 |
| Sonnet 5 | $0.00271 | $0.00271 |
| Haiku 4.5 | $0.00136 | $0.00136 |
Grade A, and why
Torus-Framework CLAUDE.md scanned grade A with 0 findings 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Torus-Framework
TOOLSHED (single MCP gateway)
All tools route through toolshed: run_tool(server, tool, args)
- memory: search_knowledge, get_memory, remember_this, fuzzy_search, record_attempt, record_outcome, query_fix_history, health_check, agent_coordination
- torus-skills: list_skills, invoke_skill, search_skills, self_improve, skill_usage, skill_health, record_outcome, trigger_evolution, capture_skill, skill_lineage
- search: terminal_history_search, transcript_context
- web-search: web_search
- indexer: code_search, code_callers, code_readers, code_path, code_blast_radius, code_hubs, code_clusters, code_cluster_detail
- Discovery:
list_tools(group="memory")to see tools in a group
INDEX FIRST (for code searches)
BEFORE exploring code: Read .claude/GRAPH_REPORT.md in the project root for hubs, clusters, and file map.
- For targeted lookups:
run_tool("indexer", "code_search", {"project": "<dir-name>", "query": "..."}) code_callers/code_readersfor call chains,code_blast_radiusfor impact- If index returns hits, Read those files directly — do NOT grep/find/glob
MEMORY FIRST (Non-Negotiable)
BEFORE building/fixing ANYTHING: run_tool("memory", "search_knowledge", {"query": "..."})
-
0.5: use directly | 0.2-0.5: get_memory(id) to verify | <0.2: treat as unknown AFTER any fix/decision/failed-approach/preference:
run_tool("memory", "remember_this", {"content": "...", "tags": "..."})For errors: use Causal Chain (below) then remember_this()
THE LOOP (mandatory — do not skip steps)
memory → brainstorm → writing-plans → tests → implement → prove → review → commit
- Write failing tests BEFORE implementation — define "correct" before coding
- All steps are toolshed skills:
run_tool("torus-skills", "invoke_skill", {"name": "brainstorm"}), etc. - Do NOT use EnterPlanMode — brainstorm replaces it
- For quick fixes: memory → fix → test → commit (skip brainstorm/writing-plans)
SKILLS (via toolshed MCP — NOT the built-in Skill tool)
NEVER use the Skill tool for framework skills (brainstorm, commit, wrap-up, etc.)
The built-in Skill tool is ONLY for: update-config, simplify, loop, schedule, claude-api, keybindings-help.
All framework skills route through: run_tool("torus-skills", "invoke_skill", {"name": "..."})
Discovery: run_tool("torus-skills", "search_skills", {"query": "..."})
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 · 76 lines · 1,357 tokens per session scan A 21712abe4cd3
Torus-Framework CLAUDE.md is an instructions file published in the GitHub repository OZmasterAI/Torus-Framework (5 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 1,357 tokens to every session, about $0.0068 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
spellbook AGENTS.md
Instructions for majiayu000/spellbook, covering spellbook agent contract, routing, scope rules, threads long-run guardrails and validation.
open-supermarkets AGENTS.md
Instructions for abracadabra50/open-supermarkets, covering agent integration guide, supported frameworks, quick integration, 1. add as skill and 2. agent calls commands.
helm copilot-instructions.md
Instructions for PetePeter/helm, covering gamepad-cli-hub — copilot instructions, project purpose, system overview, data flow pipeline and key controls.
fusion-skills AGENTS.md
AGENTS.md instructions for CrowdStrike/fusion-skills, covering agents.md, what this is, prerequisites, repository structure and skills ecosystem.
aeon CLAUDE.md
Instructions for aeonfun/aeon, covering aeon, how aeon works, strategy, voice and soul file hierarchy (read in this order).
ESAA-Core AGENTS.md
Instructions for elzobrito/ESAA-Core, covering agents.md — contrato operacional codex/esaa, 1. autoridade e termos, 2. cli e runner, ou configure o runner uma vez por sessão and 3. concorrência.