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/aquifer-sea/pattern8/agents-mdgit clone --depth 1 https://github.com/Aquifer-sea/pattern8What 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 | $0.00414 | $0.00414 |
| Opus 5 | $0.00207 | $0.00207 |
| Sonnet 5 | $0.00083 | $0.00083 |
| Haiku 4.5 | $0.00041 | $0.00041 |
Grade A, and why
pattern8 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.
What it actually says
AGENTS.md — P8 Global Instructions (all Agents must follow)
Pattern 8 (P8) — Agent Behavior Constraints
You are working in a project governed by the Pattern 8 (P8) framework. You must follow these rules.
Core Constraints
Before starting any task, you must:
- Check the
skills/directory to find a SKILL matching the current task - Read all files in that SKILL (SKILL.md + assets/ + references/)
- Strictly follow the pipeline steps described in SKILL.md
Five Patterns (cannot be skipped)
1. Pipeline
Tasks must be executed in the order defined in SKILL.md. No steps may be skipped.
2. Inversion
Before starting work, check every item in assets/checklist.yaml.
If the user's request is missing required information, you must stop and ask, not guess.
3. Generator
Output must strictly follow the format defined in assets/template.yaml.
No freestyle — every template field must be filled.
4. Tool Wrapper
Before executing any tool call, check references/security.yaml.
Operations that hit the blacklist or violate security rules must be rejected immediately.
5. Reviewer
After completing output, self-audit using the criteria in references/guidelines.yaml.
If non-compliant, roll back and regenerate, up to 3 retries.
If still non-compliant after 3 retries, inform the user and explain why.
Security Red Lines
Rules in references/security.yaml are absolute hard constraints that must never be violated.
Violating any security red line requires you to:
- Immediately stop the current operation
- Report the violation to the user
- Do NOT attempt to bypass or work around it
SKILL Priority
If a matching SKILL is found, you must use it. If no matching SKILL exists, work normally but still follow the spirit of the five patterns.
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 · 49 lines · 414 tokens per session scan A aeb009a1d165
pattern8 AGENTS.md is an instructions file published in the GitHub repository Aquifer-sea/pattern8 (49 stars, last pushed 5mo ago), licensed MIT. It adds 414 tokens to every session, about $0.0021 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
cyrus copilot-instructions.md
Copilot instructions for cyrusagents/cyrus: Note, there is a need to maintain the use of '--print' when running the claude exec commands because that is what makes it non-interactive.
screenshotting-gui AGENTS.md
Instructions for EvilFreelancer/screenshotting-gui, covering authoring & maintaining screenshotting-gui, what this skill is, non-negotiable invariant, layout and dependency policy.
compare-mcp CLAUDE.md
Instructions for Cristophereasygoing927/compare-mcp, covering compare-mcp, architecture, running and key decisions.
ai-prompts AGENTS.md
Instructions for liatrio-labs/ai-prompts, covering ai agent instructions, context marker, local skill scripts, standard workflow and initialize.
midplane CLAUDE.md
Instructions for midplaneai/midplane, covering claude code — midplane-cloud and skill routing (claude code).
agent-done-or-not CLAUDE.md
Instructions for mohamedzhioua/agent-done-or-not, a project described as: Catch your AI agent lying about "done" A one-file proof-gate for Claude Code, Cursor & Codex. it must verify before it can finish.