Borrowing it
Nothing to install: this file belongs to froggeric/claude-smart-approval. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/froggeric/claude-smart-approval/master/CLAUDE.mdgit clone --depth 1 https://github.com/froggeric/claude-smart-approvalWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/froggeric/claude-smart-approval/claude-md)<a href="https://agentmods.dev/instructions/froggeric/claude-smart-approval/claude-md"><img src="https://agentmods.dev/badge/instructions/froggeric/claude-smart-approval/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/froggeric/claude-smart-approval/claude-md"><img src="https://agentmods.dev/badge/instructions/froggeric/claude-smart-approval/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.02314 | $0.02314 |
| Opus 5.5 | $0.00926 | $0.00926 |
| Sonnet 5.5 | $0.00463 | $0.00463 |
| Haiku 4.5 | $0.00231 | $0.00231 |
Grade D, and why
claude-smart-approval CLAUDE.md scanned grade D with 3 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 12d 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
Deny rules (and Stage 2's auto-learned allow patterns) are read from (in order): `~/.claude/settings.json`, `~/.claude/settings.local.json`, `<git-root>/.claude/settings.json`, `<git-root>/.claude/settings.local.json`. A Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
Stage 1 log (`auto-approve.log`) records deny and fast-path allow decisions: JSONL with fields `ts`, `cmd`, `decision` ∈ deny/allow, `matched_prefix`, `type` (simple/compound). Model decisions live in `smart-approval.log Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
Stage 1 log (`auto-approve.log`) records deny and fast-path allow decisions: JSONL with fields `ts`, `cmd`, `decision` ∈ deny/allow, `matched_prefix`, `type` (simple/compound). Model decisions live in `smart-approval.log How it starts
The opening of the file, as written. The whole thing — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project
Claude Code PermissionRequest hook — an AI permission tier for Bash commands. Fires only when Claude Code is about to prompt the user (requires Claude Code ≥ 2.0.45; top-level headless claude -p sessions don't fire the event). Stage 1 parses compound commands (pipes, chains, subshells) via shfmt AST, denies anything matching the user's deny lists, then approves compounds whose every parsed segment matches an allow rule (bare rules exact, starred rules as prefixes). Everything else falls through to Stage 2, which evaluates via claude -p and emits an allow/deny decision object — or nothing, which escalates to the native prompt.
Commands
# Run all tests (requires bats, shfmt, jq)
bats test/
# Run a single test file
bats test/parsing.bats
# Lint shell scripts
shellcheck -x auto-approve.sh smart-approve.sh auto-learn.sh
# Evaluate a prompt template against the test suite
./eval-prompt.sh prompts/candidate-refined.txt --runs 3 --model haiku
# Debug a specific command's parsing
echo 'ls | grep foo' | ./auto-approve.sh parse
# Debug a full hook decision (deny gate + Stage 2 dispatch)
echo '{"tool_input":{"command":"ls | grep foo"}}' | ./auto-approve.sh --debug
# Exercise the working tree as a plugin (installed copy untouched), in a pty:
script -q /tmp/out claude --plugin-dir <repo-root> --settings <default-mode.json> "run exactly: <cmd>" </dev/null
# then read the two logs and pkill the hung TUI
Live-test gotchas: headless claude -p never fires PermissionRequest (useful only to confirm inactivity). Marker commands must have zero allow AND deny rules in all four settings layers, and commands Claude Code classifies as read-only (diff --version) never reach the hook.
CI
The workflow never auto-runs on push (it is registered, but push triggers are inert; cause unknown). Run it manually:
gh workflow run CI --ref master
gh run watch <run-id> --repo froggeric/claude-smart-approval --exit-status
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.
- 12d ago Changed · +1 lines · +154 tokens per session df7d4996e362
- 1mo ago First seen · 133 lines · 2,160 tokens per session scan D b8027c4b55e5
claude-smart-approval CLAUDE.md is an instructions file published in the GitHub repository froggeric/claude-smart-approval (4 stars, last pushed 14d ago), licensed MIT. It adds 2,314 tokens to every session, about $0.0093 per session on Opus 5.5. A static security scan graded it D with 3 findings (reads agent configuration directories, recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
vscode model-prompts.instructions.md
Model-specific prompt authoring and registry guidelines.