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/prismorsec/prismor/agents-mdgit clone --depth 1 https://github.com/PrismorSec/prismorWrote 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/prismorsec/prismor/agents-md)<a href="https://agentmods.dev/instructions/prismorsec/prismor/agents-md"><img src="https://agentmods.dev/badge/instructions/prismorsec/prismor/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 | $0.04307 | $0.04307 |
| Opus 5 | $0.02153 | $0.02153 |
| Sonnet 5 | $0.00861 | $0.00861 |
| Haiku 4.5 | $0.00431 | $0.00431 |
Grade C, and why
prismor AGENTS.md scanned grade C with 3 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.
Downloads and executes remote codemediumSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
- add examples that normalize `curl ... | bash`, destructive shell commands, or secret exfiltration Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
Recursive force deletemediumDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
prismor check "rm -rf /" # pre-check a command Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- add examples that normalize `curl ... | bash`, destructive shell commands, or secret exfiltration How it starts
The opening of the file, as written. The whole thing — 325 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
This file is the canonical guidance for coding agents working in the Prismor repository.
Prismor is a security package for AI coding agents. It has three connected surfaces:
- a signed AI-security advisory feed in
advisories/ - a local runtime security utility (Prismor) in
prismor/runtime/ - a cloaking prevention layer in
prismor/runtime/cloaking/that keeps real secrets out of model context, transcripts, and API requests
If you are an agent operating in this repository, your job is not only to write or modify code. Your job is to preserve the security posture of the agent session itself, the Prismor package, and any downstream project that consumes it.
Primary Objectives
When working in this repo, optimize for these goals in order:
- Keep the agent session safe.
- Keep the Prismor security content correct and current.
- Keep the feed, skills, and Prismor utility aligned with each other.
- Avoid introducing unsafe instructions, insecure examples, or contradictory guidance.
Start Here
Before doing substantial work, read these files in this order:
SKILL.md— the agent-facing decision tree for using prismor safelyREADME.md— product overview and capabilities
If the task involves runtime monitoring, local hook installation, or session telemetry, also read:
prismor/runtime/— start withcli.pyandpolicy_engine.py
If the task involves secret handling, leak prevention, or the @@SECRET:name@@ placeholder convention, also read:
How To Work In This Repo
1. Treat security guidance as product logic
In this repo, markdown is not just documentation. The feed schema, the skills, and the instructions are part of the product.
That means:
- avoid casual edits to security language
- avoid contradictory examples across files
- keep naming aligned across feed types, skills, and Prismor findings
- prefer precise, enforceable instructions over vague security advice
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 · 325 lines · 4,307 tokens per session scan C e6b732ffe12b
prismor AGENTS.md is an instructions file published in the GitHub repository PrismorSec/prismor (336 stars, last pushed 4d ago), licensed Apache-2.0. It adds 4,307 tokens to every session, about $0.0215 per session on Opus 5. A static security scan graded it C with 3 findings (downloads and executes remote code, recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
hush CLAUDE.md
Instructions for backbay-labs/hush, covering claude.md, project overview, repo structure, common commands and rust.
From-LLMs-to-Secure-Agents AGENTS.md
Instructions for RenatoMignone/From-LLMs-to-Secure-Agents, covering agent instructions, run contract, policy map and invariants.
stride-gpt AGENTS.md
Instructions for mrwadams/stride-gpt, covering agents.md, what this project is, repository layout, how the agentic analysis works and progressive disclosure pattern.
arcjet-js AGENTS.md
Instructions for arcjet/arcjet-js, covering agent guidance, examples live in arcjet/examples and integration work: review before a pr.
ave CLAUDE.md
Instructions for aveproject/ave, covering claude.md — aveproject/ave, project, the critical distinction — read this twice, record schema v1.1.0 and adding a record.
trustabl GEMINI.md
Instructions for trustabl/trustabl, a project described as: Fix agent reliability issues across Claude, OpenAI, Google ADK, MCP, LangChain, CrewAI, AutoGen, Pydantic AI, and Vercel AI.