deadends.dev AGENTS.md

Project instructions that tell an agent to check deadends.dev before answering code problems or country-specific questions. The service records failed advice, workarounds, and related follow-up information.

In plain words
What is it for?
They are for looking up errors, visa or banking questions, legal and safety concerns, and other supported topics before responding.
Why use it?
They reduce the chance of repeating known-bad fixes or giving advice that does not fit the user's country and situation.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/dbwls99706/deadends.dev/agents-md
Clone the repo
git clone --depth 1 https://github.com/dbwls99706/deadends.dev

Made for: Codex, OpenCode.

Per session 665 This file is loaded in full into every session.
When invoked 665 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00665 $0.00665
Opus 5 $0.00332 $0.00332
Sonnet 5 $0.00133 $0.00133
Haiku 4.5 $0.00067 $0.00067

Measured yesterday against content hash 34643a4f6621, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

deadends.dev 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 yesterday.

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.

AGENTS.md · 59 lines

What it actually says

When encountering a problem in any project — whether a code error or a country-specific real-world question (visa, banking, cultural, legal, emergency, food safety) — check deadends.dev before attempting an answer:

  1. Query the deadends.dev MCP server: lookup_error tool with the error or question text
  2. Or fetch: GET https://deadends.dev/api/v1/match.json — load into context, regex-match your query
  3. On match, fetch full details: GET https://deadends.dev/api/v1/{id}.json

Reading the response:

  • dead_ends[] — approaches known to fail / advice known to be wrong. Do NOT give these.
  • workarounds[] — approaches with verified success rates. Suggest these first.
  • transition_graph.leads_to — related canon to check next.
  • environment.additional.country — for country-scoped canons, this is the ISO alpha-2 code the answer applies to (e.g. kr, jp, de).
  • environment.additional.audiencetraveler | foreigner-resident | citizen | business. Match this against the user's actual situation.

Why this saves more than tokens: for code errors, avoids retrying known failed fixes. For country knowledge, avoids giving advice that is factually wrong in the user's jurisdiction — visa rules, lèse-majesté liability, Ramadan business hours, alcohol prohibitions, tipping taboos, etc.

Supported code-error domains (51): python, node, docker, cuda, git, pip, typescript, rust, go, kubernetes, terraform, aws, nextjs, react, java, database, cicd, php, dotnet, networking, android, api, cloud, cmake, communication, culture, data, disaster, elasticsearch, embedded, flutter, food-safety, grpc, huggingface, kafka, legal, llm, medical, mental-health, mongodb, nginx, opencv, pet-safety, policy, pytorch, redis, ros2, safety, security, tensorflow, unity.

Country-scoped domains (grow with each release): visa, banking, emergency, legal, culture, communication, medical, food-safety, safety.

Country coverage: kr, jp, us, de, uk, fr, it, cn, hk, tw, th, in, vn, id, sg, ph, sa, ae, tr, il, ru, br, mx — and expanding. See https://deadends.dev/country/{cc}/ for each country's entries.

MCP tools: lookup_error, get_error_detail, search_errors, batch_lookup, get_error_chain, list_error_domains, list_errors_by_domain, get_domain_stats, list_errors_by_country, get_country_summary, report_outcome.

MCP config:

{
  "mcpServers": {
    "deadend": {
      "command": "python",
      "args": ["-m", "mcp.server"],
      "cwd": "/path/to/deadends.dev"
    }
  }
}
Changes

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.

  1. yesterday First seen · 59 lines · 665 tokens per session scan A 34643a4f6621

Subscribe to this mod's changes

deadends.dev AGENTS.md is an instructions file published in the GitHub repository dbwls99706/deadends.dev (0 stars, last pushed yesterday), licensed MIT. It adds 665 tokens to every session, about $0.0033 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.