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/zimalabs/code-decisions/claude-mdgit clone --depth 1 https://github.com/zimalabs/code-decisionsWhat 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 | $0.01229 | $0.01229 |
| Opus 5 | $0.00615 | $0.00615 |
| Sonnet 5 | $0.00246 | $0.00246 |
| Haiku 4.5 | $0.00123 | $0.00123 |
Grade A, and why
code-decisions 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 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 — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
What This Is
Decision memory for Claude Code — captures why choices were made so future sessions inherit context instead of repeating debates. Decisions are markdown files in .claude/decisions/ (committed to the repo, shared via git). Hooks and skills handle everything automatically.
Commands
uv sync # install dev deps (first time only)
make check # ruff + mypy + shellcheck + pytest
make dev # symlink plugin into Claude Code cache
Key Concepts
- Decisions extend Claude Code's memory schema with:
name,description,date,tags,affects - YAML frontmatter with
---delimiters (not TOML) for decision files. Block scalars (|,>) are rejected. - All decisions are lightweight — frontmatter, title, and lead paragraph. No weight distinction.
- DecisionStore = file-based store with FTS5 derived search index. Progressive loading: index stores summaries, full files read on demand.
- Zero-config, advise-only. No strictness settings, no config file. The plugin just nudges — it never blocks or enforces.
Storage
Decisions live at .claude/decisions/{slug}.md in the repo — committed to git and shared with the team. The FTS5 index lives at ~/.claude/projects/{project-key}/.decisions/.
Architecture Rules
Architecture lives in decision files under .claude/decisions/ — see .claude/rules/decisions.md for the full index.
- Markdown is source of truth. SQLite FTS5 is a derived index. Delete it and everything rebuilds.
- Git is the history layer. Edit decisions in place, delete when obsolete. Git tracks the evolution.
- Skill = intent + behavior, hooks = correctness, CLI = data access.
/decisionskill routes intent. Hooks enforce validation and inject context. CLI provides search, stats, and git operations. - Pure methods on DecisionStore. No side effects at import time.
- Plugin is stdlib only. Zero external dependencies in
src/. - No confirmation prompts on capture. Write decisions directly — agents must capture without human approval gates.
/decision undois the safety net.
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 · 91 lines · 1,229 tokens per session scan A a4fbba6d149c
code-decisions CLAUDE.md is an instructions file published in the GitHub repository zimalabs/code-decisions (5 stars, last pushed 1mo ago), licensed MIT. It adds 1,229 tokens to every session, about $0.0061 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
cli CLAUDE.md
Instructions for archgate/cli, covering claude.md, technology stack, commands, validation gate and git hooks (git 2.54+).
mcp-adr-analysis-server copilot-instructions.md
Instructions for tosin2013/mcp-adr-analysis-server, covering mcp adr analysis server - ai coding agent instructions, project overview, critical technical conventions, esm-only module system and typescript configuration.
nullius CLAUDE.md
Instructions for armanfatemi/nullius, covering nullius, commands, known-environmental test failures, evidence anchors — the house rule and dogfooding gates (ci runs all of these).
skilldrop AGENTS.md
Instructions for sananthanarayan/skilldrop, covering agents.md — guidance for autonomous coding agents, repo in one paragraph, golden rules, verified commands (do not invent variants) and consistency lint — run from the repo root before committing.
agent-skills AGENTS.md
Instructions for stark-ai-de/agent-skills, covering agent instructions and rules.
skilldrop CLAUDE.md
Instructions for sananthanarayan/skilldrop, covering claude.md, dogfooding a skill while you build it, then in a new claude code session in any repo, ${claudeskilldir} semantics and project settings.