Claude Code TipsBLOCK
guidance for coding with agents in production, from pet projects to startups and big tech
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
this is my evidence backed guide to working with coding agents, whether you are a student having an existential crisis, a startup founder working with real money for yourself and other people, or an engineer at one of the biggest technology companies in the world. the guidance is organized around the scale and consequences of the work.
read the handbook or go directly to a principal guide:
- codex
- claude code
- grok
- start here
- choosing an agent setup
- how coding agents got here
- where this comes from
what this helps you decide
the handbook separates the surface where you steer work, the harness that runs the agent loop, the model that supplies inference, and the orchestration used for parallel work. it also covers repository instructions, permissions, review, evidence, hardware, and the operating costs that appear after the first demo.
evidence principle
tested: reproduced in a named environment and version, with the tester and limits statedofficial source: confirmed in current primary documentation or source codeanalysis: a judgment derived from stated evidenceopen question: current evidence is missing or incomplete
citations sit beside the claims they support. tested observations name their limits, while source based claims stay distinct from personal analysis.
local verification
bun install --frozen-lockfile bun run verify
built and maintained by [ani potts](https://anipo
41255d868c47OBSERVED · 2026-10-08Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add cc --env GITHUB_TOKEN=${GITHUB_TOKEN} -- npx -y [email protected]{
"mcpServers": {
"cc": {
"command": "npx",
"args": [
"-y",
"[email protected]"
],
"env": {
"GITHUB_TOKEN": "${GITHUB_TOKEN}"
}
}
}
}Exposed tools (5)
4 read · 1 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
inspect_source | read | Returns the public title, publisher, URL, and evidence type for one source referenced by the handbook. |
list_handbook_pages | write | Lists canonical public coding agent tips pages with route, update, evidence, and source provenance. |
open_handbook_page | read | Navigates to a validated canonical handbook route and optional heading on the current site. |
read_handbook_section | read | Reads one public section by canonical route and heading anchor, returning bounded text and source provenance. |
search_handbook | read | Searches canonical public handbook text locally and returns bounded matches with provenance and normal URLs. |
Trust audit
BLOCKgrade D · trust 67/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | WARN |
| L2 | Instruction surface (what it tells the agent) | FAIL |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (7 observation(s))
- Network
- declared (6 observation(s))
- Shell
- declared (5 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- found
Findings (25)
tool until restart) — use the bash fast path below. Don't tell the user
NEVER write to the database. NEVER show SQL to the user. If a query fails, describe what data was unavailable, not which table was missing.
"requests to the human instead of acting. Exfil guard refuses paths under " +
const origin = `http://127.0.0.1:${previewPort}`;return { origin: `https://127.0.0.1:${server.address().port}`, close };const origin = `http://127.0.0.1:${previewPort}`;const origin = `http://127.0.0.1:${port}`;result = mine.sanitize_string("token ghp_abcdefghijklmnopqrstuvwxyz0123456789")text = "sk-longkeyhere12345678901234 and ghp_anotherlongkey1234567890123456789012"
.gitleaks.toml
.markdownlint.jsonc
.gitleaks.toml
plugins/lore/AGENTS.md
import { buildEditorialInventory } from '../../scripts/lib/editorial-inventory.mjs';import { buildAgentIndex } from '../../agent-index.mjs';import { contentKeyForRoute } from '../../agent-catalog.mjs';import { AGENT_CATALOG_VERSION } from '../../agent-index-version.mjs';const registry = JSON.parse(readFileSync(new URL('../../editorial/sources.json', import.meta.url), 'utf8'));- Implementation: http://127.0.0.1:4330/, desktop light theme, scroll position 0, menus closed.
@modelcontextprotocol/sdk, zod, zod-to-json-schema, typescript, @types/bun
helper, silently. Don't tell the user it's a "fallback."
public/media/guides/claude-terminal-recording.mp4
an access layer can keep credential values out of prompts and supply them to
output now tunes by the active effort level read from `process.env.CLAUDE_EFFORT`:
**runtime:** [bun](https://bun.sh) 1.1+. install with `curl -fsSL https://bun.sh/install | bash` (~15s, one-time). bun replaces node + tsx + better-sqlite3 with a single binary; cold-start is ~13× fas
Gates applied: instruction_override, no_behavioural_pass.
41255d868c47full audit observations/trust-audit/mcp-server/anipotts__claude-code-tips.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-08 | 41255d868c47 | BLOCK | D | 67 | first audit |
Questions
What is the Claude Code Tips MCP server?
guidance for coding with agents in production, from pet projects to startups and big tech
What tools does Claude Code Tips expose?
5 in total: 4 read-only, 1 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is Claude Code Tips safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (67/100) and found 2 critical or high issues in the source. Each one is listed on this page with the file and line it is on.
What credentials does Claude Code Tips need?
It reads GITHUB_TOKEN from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does Claude Code Tips run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as cc at 3.8.0.
How current is this page?
The grade is for one exact copy of the source (41255d868c47), read on 2026-10-08. The repository is watched and re-audited when it changes.