mikefullerton/llm-integration-policies
skills/llm-integration-policies/SKILL.md
Use when a file imports an LLM SDK (anthropic, openai, langchain, etc.), when scaffolding LLM features, or when placing Claude-specific files in a repo. Enforces the "do not assume Claude" rule, the /claude directory convention, Graphify opt-in, and the rules-vs-behavioral-instructions split.
development
Updated Apr 21, 2026
$ install --global
skillsauthnpx skillsauth add mikefullerton/catherding llm-integration-policiesInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 22, 2026, 2:49 AM182.9s1 file scanned
SKILL.md
- name:
- llm-integration-policies
- description:
- Use when a file imports an LLM SDK (anthropic, openai, langchain, etc.), when scaffolding LLM features, or when placing Claude-specific files in a repo. Enforces the "do not assume Claude" rule, the /claude directory convention, Graphify opt-in, and the rules-vs-behavioral-instructions split.
LLM Integration — MANDATORY
Do not assume Claude
- Do not assume Claude is the LLM being used in a repo unless the repo is explicitly Claude-specific (e.g. a Claude skill, plugin, or agent).
- If the repo is not Claude-specific, any Claude-related files go in a top-level
/claudedirectory — not scattered across the repo.
Graphify opt-in
Every repo should be opted into Graphify (for every LLM that supports it):
- Graphify installed on the machine:
pip install graphifyy && graphify install graphify-out/added to.gitignore(generated output, not committed)- A
.graphifyignoreif any directories should be excluded (same syntax as.gitignore) - Run
/graphifyin a session to generate the initial graph
Rules vs Behavioral Instructions
Two different things. Pick the right one:
- General policies — apply regardless of which LLM or tool is in use. Written in plain language for anyone (human or AI) to follow. Live in
/policies/. - Claude-specific behavioral instructions — tell Claude how to behave ("when X happens, do Y"). Live in
.claude/CLAUDE.md(project scope) or/claude/rules/(global scope). Written as directives to Claude specifically.
Heuristic: If a rule would make sense for a human developer to follow, it's a general policy. If it's "Claude, remember to...", it's a behavioral instruction.
Claude Extensions
Extensions (skills, rules, agents, MCP integrations, commands, plugins, hooks) are scoped by who they affect:
- Global — affect Claude across all projects. Live in
/claude/<type>/(e.g./claude/skills/,/claude/rules/). Must have install/uninstall scripts in/setup/. - Repo-local — only affect Claude inside this repo. Live in
.claude/(e.g..claude/skills/,.claude/CLAUDE.md). No install scripts needed — Claude Code picks them up automatically.
Reference
Full rationale in ~/projects/active/catherding/policies/setup/repo-organization/llm-file-layout.md. See INDEX.md for related policies.
Related Skills
mikefullerton/general-principles
development
VerifiedTrustedCommunity
Use whenever writing, modifying, refactoring, or reviewing code, or when making any design decision — including small ones. Loads the 21 cookbook principles (simplicity, yagni, fail-fast, explicit-over-implicit, separation-of-concerns, design-for-deletion, etc.) as heuristics for judgment calls about code shape, module boundaries, and what to build. Applies across all languages and project types.
SKILL.mdUpdated Apr 22, 2026
mikefullerton/yolo
tools
VerifiedTrustedCommunity
Toggle per-session YOLO mode (auto-approve permissions with configurable deny list). Use when --dangerously-skip-permissions is broken. /yolo on, /yolo off, /yolo install, /yolo uninstall, /yolo configure, /yolo status
SKILL.mdUpdated Apr 21, 2026
mikefullerton/setup-and-install-scripts
development
VerifiedTrustedCommunity
Use when authoring or modifying install/uninstall scripts, or when creating a /setup directory for a repo that requires developer setup steps. Enforces the /setup layout, install/uninstall naming, and supporting-files location.
SKILL.mdUpdated Apr 21, 2026
mikefullerton/new-repo-scaffold
testing
VerifiedTrustedCommunity
Use when creating a new repo (git init, gh repo create, scaffolding a new project) or when a repo is missing README/LICENSE/.gitignore/.claude/CLAUDE.md. Enforces the "every repo must have" checklist and documentation layout.
SKILL.mdUpdated Apr 21, 2026