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/subvertnormality/paranoia/agents-mdgit clone --depth 1 https://github.com/subvertnormality/paranoiaWhat 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 | $0.08635 | $0.08635 |
| Opus 5 | $0.04318 | $0.04318 |
| Sonnet 5 | $0.01727 | $0.01727 |
| Haiku 4.5 | $0.00864 | $0.00864 |
Grade A, and why
paranoia 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 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.
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 — 548 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repository agent instructions
These instructions govern work in paranoia-local. User instructions take precedence but do not silently expand scope.
Freeze proportionate stakes
Before any convergence loop, state one concrete, stable operating model: deployment, operators, trusted actors, untrusted inputs, active adversary capabilities, concurrency, network boundary, expected files/claims/rounds/latency, consequences of false clear versus false block, and explicit exclusions. Avoid vague labels such as “production”, “malicious repository”, or “high stakes”. Distinguish untrusted static bytes from repository-selected code execution, a hostile local process racing paths, a compromised OS, and hostile web content. A material stakes change requires user direction and retriage; never strengthen stakes merely to keep a review loop generating work.
Default local-tool calibration unless the task says otherwise: trusted single user and OS;
plan/repository/fetched content untrusted as data; no hostile local race; ordinary edits must
block/retry; tens to low hundreds of claims; evidence useful within minutes; false
NOT-BLOCKED, authority errors, and wrong evidence binding high impact; recoverable blocking
acceptable. Formal proof, multi-tenancy, hostile same-user races, and deliberately corrupted
state recovery are not implied.
Review the right artifact
Do a plan review to validate the design, then implement it. Once implementation begins,
paranoia convergence reviews the code branch/diff, not the old plan, unless the user
explicitly asks to reopen plan design. Never spend repeated rounds polishing a plan while
reporting implementation progress. Label every run in commentary as PLAN or CODE.
After implementation begins, do not invoke critique_plan again merely as acceptance for
plan-role mechanics unless the user explicitly authorizes that return. Use deterministic
production-handler lifecycle coverage and keep convergence on the code artifact.
Before review, update public documentation and these agent instructions. Before opening a PR, run the primary capability end to end; fake-backed tests alone do not establish usability.
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 · 548 lines · 8,635 tokens per session scan A f89ba4a6aa63
paranoia AGENTS.md is an instructions file published in the GitHub repository subvertnormality/paranoia (13 stars, last pushed 2d ago), licensed MIT. It adds 8,635 tokens to every session, about $0.0432 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
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
buildNext
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
spec-kit AGENTS.md
Instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
langchain AGENTS.md
Instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.