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/managedcode/geminisharpsdk/agents-mdgit clone --depth 1 https://github.com/managedcode/GeminiSharpSDKWhat 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 | $0.04134 | $0.04134 |
| Opus 5 | $0.02067 | $0.02067 |
| Sonnet 5 | $0.00827 | $0.00827 |
| Haiku 4.5 | $0.00413 | $0.00413 |
Grade A, and why
GeminiSharpSDK 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 yesterday.
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 — 252 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Project: ManagedCode.GeminiSharpSDK Stack: .NET 10, C# 14, TUnit, GitHub Actions, NativeAOT, NuGet, Gemini CLI integration
Follows MCAF
Conversations (Self-Learning)
Learn the user's habits, preferences, and working style. Extract rules from conversations, save to "## Rules to follow", and generate code according to the user's personal rules.
Update requirement (core mechanism):
Before doing ANY task, evaluate the latest user message.
If you detect a new rule, correction, preference, or change -> update AGENTS.md first.
Only after updating the file you may produce the task output.
If no new rule is detected -> do not update the file.
When to extract rules:
- prohibition words (never, don't, stop, avoid) or similar -> add NEVER rule
- requirement words (always, must, make sure, should) or similar -> add ALWAYS rule
- memory words (remember, keep in mind, note that) or similar -> add rule
- process words (the process is, the workflow is, we do it like) or similar -> add to workflow
- future words (from now on, going forward) or similar -> add permanent rule
Preferences -> add to Preferences section:
- positive (I like, I prefer, this is better) or similar -> Likes
- negative (I don't like, I hate, this is bad) or similar -> Dislikes
- comparison (prefer X over Y, use X instead of Y) or similar -> preference rule
Corrections -> update or add rule:
- error indication (this is wrong, incorrect, broken) or similar -> fix and add rule
- repetition frustration (don't do this again, you ignored, you missed) or similar -> emphatic rule
- manual fixes by user -> extract what changed and why
Strong signal (add IMMEDIATELY):
- swearing, frustration, anger, sarcasm -> critical rule
- ALL CAPS, excessive punctuation (!!!, ???) -> high priority
- same mistake twice -> permanent emphatic rule
- user undoes your changes -> understand why, prevent
Ignore (do NOT add):
- temporary scope (only for now, just this time, for this task) or similar
- one-off exceptions
- context-specific instructions for current task only
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.
- yesterday First seen · 252 lines · 4,134 tokens per session scan A cbe8edcbb89f
GeminiSharpSDK AGENTS.md is an instructions file published in the GitHub repository managedcode/GeminiSharpSDK (2 stars, last pushed 4mo ago), licensed MIT. It adds 4,134 tokens to every session, about $0.0207 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
security GEMINI.md
Instructions for gemini-cli-extensions/security, covering standard operating procedures: security analysis guidelines, persona and guiding principles, skillset: permitted tools & investigation, skillset: sast vulnerability analysis and 1.1. hardcoded secrets.
jules GEMINI.md
Instructions for gemini-cli-extensions/jules, covering jules extension, command: /jules, activation and response, starting a jules task and jules-related queries.
self-command GEMINI.md
Instructions for stevenAthompson/self-command, covering self command extension, tools, 1. selfcommand, 2. geminisleep and 3. watchlog.
run-long-command GEMINI.md
Instructions for stevenAthompson/run-long-command, covering run long command extension, usage, tool signature, example and how it works.
AionUi AGENTS.md
Instructions for iOfficeAI/AionUi, covering aionui - project guide, code conventions, file & directory structure, naming and ui library & icons.
ccmanager CLAUDE.md
Instructions for kbwo/ccmanager, covering repository guidelines, project structure & module organization, build, test & development commands, coding style & naming conventions and testing guidelines.