deadends.dev CLAUDE.md

Project instructions for deadends.dev, a website and database of coding errors and country-specific real-world problems. They describe the project structure and its validate, generate, build, and test workflow.

In plain words
What is it for?
Use them when working on deadends.dev, especially when updating its error entries, generation pipeline, or static website build.
Why use it?
They give a coding agent the project context and expected commands, reducing the chance of changing the wrong part or skipping required checks.

Instructions file

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/claude-md
Clone the repo
git clone --depth 1 https://github.com/dbwls99706/deadends.dev
Per session 3,923 This file is loaded in full into every session.
When invoked 3,923 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 $0.03923 $0.03923
Opus 5 $0.01962 $0.01962
Sonnet 5 $0.00785 $0.00785
Haiku 4.5 $0.00392 $0.00392

Measured yesterday against content hash 0c76afe31f9e, 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 CLAUDE.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.

CLAUDE.md · 322 lines

How it starts

The opening of the file, as written. The whole thing — 322 lines — stays where its author put it; the contents beside it link to each section on GitHub.

deadends.dev

Structured failure knowledge database for AI agents - covering both code errors AND country-specific real-world dead ends. 2,393+ ErrorCanon JSON entries across 54 domains:

  • 51 code-error domains (2,089 entries): python, node, docker, cuda, git, pip, typescript, rust, go, kubernetes, terraform, aws, nextjs, react, java, database, cicd, php, dotnet, networking, android, api, cloud, cmake, elasticsearch, embedded, flutter, grpc, huggingface, kafka, llm, mongodb, nginx, opencv, pytorch, redis, ros2, tensorflow, unity, + quirk domains (communication, culture, data, disaster, food-safety, legal, medical, mental-health, pet-safety, policy, safety, security).
  • 3 new country-scoped domains (visa, banking, emergency) plus existing non-code domains extended with country-scoped canons. ID format {domain}/{slug}/{country} where country is an ISO 3166-1 alpha-2 code (kr, jp, us, de, uk, fr, cn, hk, tw, th, in, vn, id, sg, ph, sa, ae, tr, il, ru, br, mx, ...). 52 countries, 300+ country-scoped entries as of v0.10.

Country-canon rationale: coding dead ends are largely solved by modern LLMs. Country-specific rules (visa caveats, jurisdictional laws, cultural taboos, food safety, emergency numbers, housing contract quirks) are where generic AI advice is most often wrong. The expansion fills this gap with primary-source-cited structured entries.

Architecture

data/
  canons/              # ErrorCanon JSON files (source of truth, 2393+ files)
    {domain}/          # One directory per domain (54 domains)
      {slug}_{env}.json          # Flat-file format (code: env = runtime-os)
      {slug}/{env}.json          # Directory-style format
                                  # For country canons: env = ISO alpha-2 code (kr, jp, ...)
  environments/        # Environment matrix (env_matrix.json)
  graph/               # Error transition graph data
  signatures/          # Error signature data

generator/
  __init__.py
  build_site.py        # Static site builder (Jinja2 templates → site/)
  bulk_generate.py     # Generate canon JSON from seed definitions (current version)
                       # Historical versions v2-v16 live under archive/bulk_generate_history/
  collect_evidence.py  # Evidence collection utilities
  collect_signatures.py # Signature collection utilities
  domains.py           # Domain constants: KEYWORD_MAP, DOMAIN_DISPLAY_NAMES, suggest_domains()
  generate_canons.py   # Canon generation logic
  generate_pairs.py    # Pair generation utilities
  lookup.py            # Programmatic error lookup SDK (lookup, lookup_all, search, batch_lookup)
  ping_search_engines.py  # Search engine ping on deploy
  pipeline.py          # Unified pipeline: validate → generate → build → test
  reverify.py          # Assigns each content cycle a disjoint bucket of aging
                       # canons to re-verify (see 'Re-verifying Aging Canons')
  schema.py            # ErrorCanon JSON Schema (ERRORCANON_SCHEMA)
  submit_indexnow.py   # IndexNow submission on deploy
  templates/           # Jinja2 HTML templates
    index.html         # Homepage
    domain.html        # Domain listing page
    error_summary.html # Error summary page (groups environments)
    page.html          # Individual error page (per environment)
    search.html        # Search page
  validate.py          # Validation: schema, business rules, HTML, cross-refs, staleness

mcp/
  server.py            # MCP server (JSON-RPC over stdio) - 11 tools for AI agents

api/
  mcp.py               # Vercel serverless MCP endpoint

site/                  # Generated static site output (DO NOT edit directly)
tests/                 # pytest test suite

Read the full file on GitHub · 322 lines

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 · 322 lines · 3,923 tokens per session scan A 0c76afe31f9e

Subscribe to this mod's changes

deadends.dev CLAUDE.md is an instructions file published in the GitHub repository dbwls99706/deadends.dev (0 stars, last pushed yesterday), licensed MIT. It adds 3,923 tokens to every session, about $0.0196 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.