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/supa-media/context/claude-mdgit clone --depth 1 https://github.com/Supa-Media/contextWhat 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 | $0.19425 | $0.19425 |
| Opus 5 | $0.09713 | $0.09713 |
| Sonnet 5 | $0.03885 | $0.03885 |
| Haiku 4.5 | $0.01943 | $0.01943 |
Grade B, and why
context CLAUDE.md scanned grade B with 2 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
places to be careful: the file (`~/.claude/settings.json`, `~/.codex/hooks.json`, Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
`{"results":[{id,title,text,url}]}`; `fetch(id)` → `{id,title,text,url, How it starts
The opening of the file, as written. The whole thing — 1,434 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Instructions — Context
Free your context. Share your context.
Context is one MCP endpoint a person adds to every AI client, backed by a
markdown bucket they own. Read README.md first for the product shape.
Non-negotiables
These are the product, not implementation details. If a task seems to require breaking one, stop and say so rather than working around it.
1. The customer owns the storage
- Canonical Markdown and attachments live in a bucket the customer owns (R2, S3, B2, any S3-compatible endpoint). We are a tenant in it.
- The control plane is never the canonical store. It holds metadata only: accounts, workspaces, storage bindings, grants, audit. Never note content.
- A customer can revoke our storage credential without asking us first, and keep a complete, usable context.
- Credentials never live in Markdown, in the customer's bucket, in logs, in URLs, or on a device. Encrypted at rest in the control plane, decrypted only in the gateway at request time.
2. Tenancy is bucket-level, never prefix-level
Do not namespace keys inside a customer's bucket. No tenants/<id>/, no
workspaces/<slug>/. A note lives at 1-projects/foo.md, full stop.
This is load-bearing. An existing brain — including Seyi's live bucket, which has been running since August — must connect to Context and work unchanged, with zero migration. Users also sync the same bucket to Obsidian via Remotely Save; rewriting keys breaks that. One workspace maps to one bucket (optionally one bucket + a fixed root prefix chosen by the customer, applied at the adapter boundary and invisible above it).
3. Plain files stay canonical
- Markdown stays portable and human-readable.
- Search indexes, caches, and embeddings are disposable derivatives, rebuildable from the files. Never the only copy of anything.
- The on-bucket layout is a stable format, not an internal detail:
index.mdandprivacy.mdat root,.history/,.audit/, PARA folders. Treat changes to it as breaking changes.
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 · 1,434 lines · 19,425 tokens per session scan B cc339c9129a3
context CLAUDE.md is an instructions file published in the GitHub repository Supa-Media/context (2 stars, last pushed 2d ago), licensed MIT. It adds 19,425 tokens to every session, about $0.0971 per session on Opus 5. A static security scan graded it B with 2 findings (reads agent configuration directories, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
mcp-documentation-server copilot-instructions.md
Instructions for andrea9293/mcp-documentation-server, covering mcp documentation server - ai coding agent instructions, architecture overview, core components, server entry point (src/server.ts) and embedding providers (src/embedding-provider.ts).
seekstone CLAUDE.md
Instructions for shaqmughal/seekstone, covering claude.md, what this repo is, commands, the harness itself (run after npm install) and architecture.
data-olympus AGENTS.md
Instructions for knaisoma/data-olympus, covering data olympus, project overview, architecture, development workflow and code standards.
markdown-vault-mcp CLAUDE.md
Instructions for pvliesdonk/markdown-vault-mcp, a project described as: Generic markdown collection MCP server with FTS5 + semantic search, frontmatter-aware indexing, and incremental reindexing.
data-olympus copilot-instructions.md
Instructions for knaisoma/data-olympus: Read and follow all instructions in AGENTS.md. Read and follow all rules in .rules/. Read and follow the global Codex and Claude Code collaboration rule in /kn-projects/company-knowledge/tooling/codex-claude-collaboration.md.
data-olympus CLAUDE.md
Instructions for knaisoma/data-olympus: Read and follow all instructions in AGENTS.md. Read and follow all rules in .rules/. Read and follow the global Codex and Claude Code collaboration rule in /kn-projects/company-knowledge/tooling/codex-claude-collaboration.md.