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/antropos17/aegis/agents-mdgit clone --depth 1 https://github.com/antropos17/AegisWrote 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/antropos17/aegis/agents-md)<a href="https://agentmods.dev/instructions/antropos17/aegis/agents-md"><img src="https://agentmods.dev/badge/instructions/antropos17/aegis/agents-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 | $0.01725 | $0.01725 |
| Opus 5 | $0.00863 | $0.00863 |
| Sonnet 5 | $0.00345 | $0.00345 |
| Haiku 4.5 | $0.00172 | $0.00172 |
Grade A, and why
Aegis 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 4d 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 — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Instructions for AI agents contributing to this codebase.
Project overview
AEGIS is an independent AI oversight layer — a desktop app that monitors AI agents running on a user's local machine. Electron 43 + Svelte 5 + Vitest.
Architecture
src/main/ Electron main process (CommonJS, require/module.exports)
src/renderer/ Svelte 5 dashboard UI (ES modules, runes)
src/shared/ Constants + agent-database.json (110 agents, 262 name signatures) + types/ (9 .ts)
rules/ 73 detection rules in 8 YAML files + _schema.json
rules/sequences/ 1 sequence correlation rule (SEQ001) in 1 sequence rule file; loaded, no consumer yet
tests/ Vitest unit tests with v8 coverage
Key modules:
src/main/scan-loop.js— core scanning pipeline (process + file + network + LLM probing)src/main/process-scanner.js— AI agent detection via process listsrc/main/anomaly-detector.js— multi-dimensional anomaly scoring (network/fs/process/baseline)src/main/llm-runtime-detector.js— local LLM runtime detection (Ollama, LM Studio HTTP probes)src/main/cli.js— CLI interface (--scan-json,--version,--help)src/main/platform/— OS abstraction (win32.js, darwin.js, linux.js)src/shared/agent-database.json— 110 known agents; theirnamesarrays hold 262 process-name signatures in totalsrc/main/rule-loader.js— loadsrules/*.yaml(73 rules, 8 categories) againstrules/_schema.json, hot-reloadsrc/shared/constants.js— ignore patterns, editor lists, AGENT_CONFIG_PATHS
Build & test
npm ci # what CI installs with — NOT npm install, see the lockfile note below
npm run build:renderer # Vite build — must succeed
npm run format:check # Prettier — must be clean (the WRITING variant is npm run format)
npm run lint # ESLint — no errors
npm run typecheck # one command, two tsc projects: tsconfig.main.json && tsconfig.renderer.json
npm run typecheck:svelte # svelte-check — reads .svelte templates, which tsc never does
npm run test:coverage # vitest run --coverage — this, not npm test, is what CI runs
npm run verify:gate # injection proof: 4 mutants must turn the witness suite red
npm run verify:seq-gate # injection proof: 4 sequence-engine mutants must turn the gate suite red
npm run counts:check # derived counts vs the numbers tracked files declare
npm audit --audit-level=high --omit=dev # production dependencies 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.
- 4d ago First seen · 96 lines · 1,725 tokens per session scan A c95b261768af
Aegis AGENTS.md is an instructions file published in the GitHub repository antropos17/Aegis (145 stars, last pushed 10d ago), licensed MIT. It adds 1,725 tokens to every session, about $0.0086 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
outline-apps AGENTS.md
AGENTS.md instructions for OutlineFoundation/outline-apps, covering welcome to the outline monorepo!, repository structure, the npm run action command, common actions and other useful commands.
outline-apps CLAUDE.md
Claude Code instructions for OutlineFoundation/outline-apps: This project uses AGENTS.md instead of CLAUDE.md.
agent-nudge AGENTS.md
AGENTS.md instructions for manazoid4/agent-nudge, a project described as: Local-first pre-action context assurance for AI coding agents: detect stale context and collisions before edits.
agent-nudge CLAUDE.md
Claude Code instructions for manazoid4/agent-nudge: Read AGENTS.md, PRODUCT.md, DESIGN.md, and docs/PROTOCOL.md before changing behavior. Agent Nudge transports structured, source-backed context deltas; it is not a shared-memory dump. Do not edit real user Claude/Codex configuration during development or tests.
vscode tree-widgets.instructions.md
Use when asked to consume workbench tree widgets in VS Code.
vscode accessibility.instructions.md
Use when implementing accessibility features, ARIA labels, screen reader support, accessible help dialogs, or keybinding scoping for accessibility. Covers AccessibleContentProvider, CONTEXTACCESSIBILITYMODEENABLED, verbosity settings, and announcement patterns.