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/cyberchitta/llm-context.py/claude-mdgit clone --depth 1 https://github.com/cyberchitta/llm-context.pyWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/cyberchitta/llm-context.py/claude-md)<a href="https://agentmods.dev/instructions/cyberchitta/llm-context.py/claude-md"><img src="https://agentmods.dev/badge/instructions/cyberchitta/llm-context.py/claude-md.svg" alt="Measured on agentmods" height="20"></a>What 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 5.1 | $0.00582 | $0.00582 |
| Opus 5 | $0.00291 | $0.00291 |
| Sonnet 5 | $0.00116 | $0.00116 |
| Haiku 4.5 | $0.00058 | $0.00058 |
Grade A, and why
llm-context.py 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 6d 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 — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Working notes (gitignored)
Tracked separately in the private working-notes repo, symlinked at _notes/;
invisible to the Grep/Glob tools (global CLAUDE.md has the rg forms).
Read directly when relevant:
_notes/field-notes.md— friction log for thelc-curate-contextskill, written from consumer sessions in other repos. Newest entry at the top._notes/implementation-log.md— what shipped from a drained field note. Entries stand on their own; the note each came from is gone. Its header carries the next entry point; this repo keeps no separate worklist.
Speculative unless stated otherwise. Do not implement from these without asking.
Field notes: _notes/field-notes.md
Draining the field notes
An observation earns a change after 2–3 sightings in separate contexts; bugs, and notes that resonate with an already-recorded candidate, promote on one. Be suspicious of frequency from a single author — one observation refined twice is still one observation.
A promoted entry is deleted, and a summary that stands alone goes in
_notes/implementation-log.md. This is the file's only retirement gate, and it
went un-run from 2026-05-17 to 2026-08-26, by which point the file was 16KB and
half of it described behaviour that no longer existed.
Where a promoted note lands:
| Observation | Target |
|---|---|
| Wrong output, crash, or a broken documented invariant | code + a regression test |
| A command or instruction the tool emits that does not work | the template, and a test over the rendered artifact |
| Rule authoring: what to select, and how to verify it | skill SKILL.md / PATTERNS.md / EXAMPLES.md |
| Rule file schema and field semantics | skill SYNTAX.md |
| CLI or MCP behaviour, flags, output routing | skill COMMANDS.md |
| A failure mode a reader will hit again | skill TROUBLESHOOTING.md |
| Consumers without skill support | lc_resources/rules/lc/ins-rule-framework.md |
| Product direction, not friction | leave in field-notes.md, or track it where it can be tested |
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.
- 6d ago First seen · 48 lines · 582 tokens per session scan A e0eef2b5528e
llm-context.py CLAUDE.md is an instructions file published in the GitHub repository cyberchitta/llm-context.py (306 stars, last pushed 10d ago), licensed Apache-2.0. It adds 582 tokens to every session, about $0.0029 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
mcp-coroot CLAUDE.md
Claude Code instructions for jamesbrink/mcp-coroot, covering claude desktop configuration for mcp coroot, quick setup, using published package (recommended), for sso/mfa users and for local development.
powerbi-mcp AGENTS.md
Instructions for sulaiman013/powerbi-mcp, covering power bi mcp server - agent guide, what this server does, golden rules, recommended workflows and prompts (guided workflows).
powerbi-mcp CLAUDE.md
Instructions for sulaiman013/powerbi-mcp, covering claude.md and developer notes.
mem0-mcp CLAUDE.md
Claude Code instructions for andresmorales07/mem0-mcp, covering claude.md, what this is, commands, manual smoke test against a running mem0 instance and architecture.
frame-check AGENTS.md
AGENTS.md instructions for lluvr/frame-check, covering agents.md, what this repo is, what goes in this repo, public canon discipline and what changes belong here.
outlook-mcp-readonly CLAUDE.md
Claude Code instructions for r-rohit/outlook-mcp-readonly, covering claude.md — outlook-readonly-mcp, what this project is, the one hard rule of this codebase, architecture and environment constraints.