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/sageox/ox/agents-mdgit clone --depth 1 https://github.com/sageox/oxWhat 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 | $0.04266 | $0.04266 |
| Opus 5 | $0.02133 | $0.02133 |
| Sonnet 5 | $0.00853 | $0.00853 |
| Haiku 4.5 | $0.00427 | $0.00427 |
Grade A, and why
ox AGENTS.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 — 300 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run ox agent prime at session start to load SageOx team context — it's fine to read the user's request first.
Human attention is the scarcest resource
Human time is more expensive than compute. Before interrupting with a question, a PR for review, or a new issue, ask yourself: could I have resolved this without a human in the loop?
Default to independent resolution:
- Can you debug the issue yourself by reading code, running tests, checking logs, or using tools? Do that.
- Is the decision reversible? If yes, make the call and note the reasoning in a commit message. The human will tell you if they disagree.
- Do you have enough context from
CLAUDE.md, past sessions, or the codebase to decide? Decide. - Is this a genuinely novel tradeoff that sets precedent? That is worth interrupting for.
Prefer fewer, denser PRs over many small ones. A reviewer context-switching across four related PRs costs more than reviewing one larger coherent PR. Split only when the units of work are independent (different subsystems, different reviewers, different merge timing).
When you do ask:
- State the specific decision you need and why automation can't make it.
- Bundle related questions into one exchange rather than a stream of individual ones.
- Propose a default so the response can be "yes" / "no" / "do something else specific" rather than open-ended.
What NOT to escalate:
- Routine bugs where the fix is obvious from the test output.
- Style/lint issues.
- Choosing between two roughly-equivalent approaches when either would be fine.
- Asking permission for reversible operations (new branches, test runs, log analysis).
The cheapest review is the one that never has to happen because the code was right the first time. The next cheapest is a single coherent PR that tells a complete story. Optimize for those.
What is ox?
ox is agentic context infrastructure for software teams. It makes architectural decisions, team knowledge, and session history automatically available to AI coworkers — so every coding session starts with the full picture, not from zero.
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 · 300 lines · 4,266 tokens per session scan A a652f09d62f3
ox AGENTS.md is an instructions file published in the GitHub repository sageox/ox (50 stars, last pushed 4d ago), licensed MIT. It adds 4,266 tokens to every session, about $0.0213 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 2026-08-30.
Other instructions, from other repositories
obsidian-mind CLAUDE.md
Claude Code instructions for breferrari/obsidian-mind, covering obsidian mind, skills & capabilities, custom slash commands, vault structure and obsidian cli.
TokenTracker CLAUDE.md
Claude Code instructions for xiufengsun/TokenTracker, covering claude.md, project shape, frequently used commands, what's where and load-bearing conventions.
obsidian-mind AGENTS.md
AGENTS.md instructions for breferrari/obsidian-mind, covering obsidian mind, hooks, commands, memory and reaching this vault from another repo.
parallel-code CLAUDE.md
Instructions for johannesjo/parallel-code, covering parallel code, stack, commands, project structure and conventions.
neo4j-skills AGENTS.md
Instructions for neo4j-contrib/neo4j-skills, covering writing skills for this repository, language style — write for agents, not humans, six design principles, skill.md spec (agentskills.io) and frontmatter.
switchback-running CLAUDE.md
Instructions for rlacombe/switchback-running, covering ultrarunning training companion, constitution, athlete data, training philosophy and core principles.