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/husker/a2acast/copilot-instructionsgit clone --depth 1 https://github.com/husker/a2acastWhat 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 | $0.00645 | $0.00645 |
| Opus 5 | $0.00322 | $0.00322 |
| Sonnet 5 | $0.00129 | $0.00129 |
| Haiku 4.5 | $0.00064 | $0.00064 |
Grade A, and why
a2acast copilot-instructions.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 — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.
a2acast — Copilot instructions
Cross-machine agent-to-agent mesh. One stdlib-only Python file (mesh.py) plus harness
integrations for Claude Code, Codex CLI, and Copilot CLI. Tests in tests/test_mesh.py.
Full rationale for every rule below is in AGENTS.md. The rules
themselves are repeated here because an instruction you have to go fetch is an
instruction that does not get followed.
Hard constraints
mesh.pyis stdlib-only. No third-party imports. CI catches this (stdlib_onlyjob, #140) and the local suite does too (StdlibOnlyTests); a third-party import fails before push. Stay vigilant — the test catches imports, not subtle misuse of a stdlib facility.- Python 3.8+. No
match, no runtimeX | Yunions, no 3.9+ stdlib. - Tests:
python3 -m unittest discover -s tests(pythonon Windows)
Security-relevant changes
Crypto, identity, pins, certs, revocation, verdicts, wire format, or anything on a #62 / #74 / #76 / #93 line.
- Never claim a property you have not tested. The PR body is a hypothesis. Three PRs in one batch on 2026-07-25 each asserted the exact property their code lacked. Write what the code does; if a guard is partial, name the case it misses.
- Ship a test that FAILS without the fix, not one that passes with it.
- A guard must be load-bearing both ways. Revert the fix → attack test fails. Disable the mechanism → legitimate-use test fails.
- Gate on the property, not a proxy. "Sender has a list" ≠ "fleet can parse this." "Name is unpinned" ≠ "name is unclaimed."
- Mixed versions are normal. New signed fields partition older peers. Default new wire behaviour OFF, gated on an explicit operator assertion of fleet capability.
- The receive path mutates trust state only behind an owner opt-in that defaults OFF.
A received frame may write identity records only via an explicit opt-in that defaults
off (#137 resolved A; #93's
rename_migration) — never the default, never the same PR. - Design-first for signed-frame-body changes, before implementing.
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 · 48 lines · 645 tokens per session scan A 61fc14ce8228
a2acast copilot-instructions.md is an instructions file published in the GitHub repository husker/a2acast (7 stars, last pushed 5d ago), licensed MIT. It adds 645 tokens to every session, about $0.0032 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
adk-go AGENTS.md
AGENTS.md instructions for google/adk-go, covering agents.md, project overview, skills, setup & core commands and definition of done.
adk-go CLAUDE.md
Claude Code instructions for google/adk-go: See AGENTS.md for project context, commands, and contribution guidelines for AI coding agents.
a2a-python AGENTS.md
Instructions for a2aproject/a2a-python, covering agents.md, technology stack & architecture, mandatory workflow and optional extras.
any-agent AGENTS.md
Instructions for mozilla-ai/any-agent, covering claude.md, where to look first, project structure & module organization, build, test, and development commands and coding style & naming conventions.
lucid-agents AGENTS.md
Instructions for daydreamsai/lucid-agents, covering lucid agents monorepo - ai coding guide, project overview, brand system, architecture overview and package dependencies.
agentarea AGENTS.md
Instructions for agentarea/agentarea, covering agents.md, where to look, structure, commands and backend (from agentarea-platform/).