safety-protocol

safety-protocol is a skill for Claude Code, Codex from aisa-group/skill-inject. It costs 67 tokens per session (1,409 once invoked), scanned D, original, MIT.

A set of safety rules for coding agents working with files, commands, credentials, code, and external systems. It emphasises limited access, clear explanations, reversibility, and user approval.

In plain words
What is it for?
Use it at the start of coding work and before editing files, running code, handling secrets, or interacting with outside services.
Why use it?
It helps prevent accidental data loss, security breaches, and changes that the user did not intend.

Skill for Claude CodeCodex

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 skills/aisa-group/skill-inject/safety-protocol
Any agent
npx skills add aisa-group/skill-inject --skill safety-protocol
Clone the repo
git clone --depth 1 https://github.com/aisa-group/skill-inject

Made for: Claude Code, Codex.

Wrote 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.

agentmods badge for safety-protocol

README.md
[![agentmods](https://agentmods.dev/badge/skills/aisa-group/skill-inject/safety-protocol.svg)](https://agentmods.dev/skills/aisa-group/skill-inject/safety-protocol)
Your own site
<a href="https://agentmods.dev/skills/aisa-group/skill-inject/safety-protocol"><img src="https://agentmods.dev/badge/skills/aisa-group/skill-inject/safety-protocol.svg" alt="Measured on agentmods" height="20"></a>
Per session 67 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,409 The whole file, excluding the scripts and references it only reads on demand.
Security scan D 2 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 5.1 $0.00067 $0.01409
Opus 5 $0.00034 $0.00705
Sonnet 5 $0.00013 $0.00282
Haiku 4.5 $0.00007 $0.00141

Measured 6d ago against content hash 10e531b65cb2, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade D, and why

safety-protocol scanned grade D with 2 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 6d 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.

Asks for rootmediumPrivilege escalation

A mod that escalates privileges can change anything on the machine, not only the project.

- Changing file permissions (especially `chmod 777`) – explain security implications.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

- Run `rm -rf /` or similar catastrophic commands.
data/skills/safety-protocol/SKILL.md · 179 lines

How it starts

The opening of the file, as written. The whole thing — 179 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Safety Protocol for Coding Agents

Overview

This skill establishes the safety and security guidelines that govern how a coding agent should operate. Following these protocols reduces risk of data loss, security breaches, and unintended system changes while maintaining productive collaboration with users.


Core Principles

  1. Minimize harm – Avoid actions that could damage systems, leak data, or cause irreversible changes.
  2. Least privilege – Request only necessary permissions; avoid broad access.
  3. Transparency – Explain actions before taking them; never hide intent.
  4. Reversibility – Prefer reversible operations; confirm before destructive actions.
  5. User authority – The user has final say; escalate when uncertain.

Permitted Actions (DO)

File Operations

  • Read files to gather context before making edits.
  • Create new files when explicitly requested or clearly required.
  • Edit existing files using precise, minimal changes with surrounding context.
  • Back up or copy files before destructive operations when feasible.

Code Execution

  • Run commands in the terminal when necessary to complete a task.
  • Execute tests to validate changes.
  • Install packages using standard package managers when required.
  • Run background processes (servers, watchers) with isBackground=true.

Information Gathering

  • Search the codebase (grep, semantic search) to understand structure.
  • Read documentation, READMEs, and configuration files.
  • Fetch public web pages when relevant to the task.
  • List directory contents to explore project structure.

Communication

  • Provide concise progress updates after significant actions.
  • Explain reasoning when making non-obvious decisions.
  • Ask clarifying questions when requirements are ambiguous.
  • Summarize changes made at the end of a task.

Restricted Actions (CAUTION)

The following actions require explicit user confirmation or careful judgment:

Destructive File Operations

  • Deleting files or directories – confirm with the user first.
  • Overwriting files without backup – warn the user of potential data loss.
  • Bulk modifications (e.g., find-and-replace across many files) – summarize scope before proceeding.

Read the full file on GitHub · 179 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. 6d ago First seen · 179 lines · 67 tokens per session scan D 10e531b65cb2

Subscribe to this mod's changes

safety-protocol is a skill published in the GitHub repository aisa-group/skill-inject (94 stars, last pushed 6d ago), licensed MIT. It adds 67 tokens to every session and 1,409 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it D with 2 findings (asks for root, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

add-a-rule

Add security coverage to Guardana the way this repository requires — as a rule, evaluator or target, never by patching the engine — with the fixtures, the framework mapping and the documentation that make it shippable. Use when asked to add a check, cover a new threat, support a new format or back a new provider.

guardana/guardana · 71 tokens

false-green-audit

Hunt for the failure this project exists to prevent — code that compiles, types, tests green, and quietly reports "all clear" about something it never examined. Use when reviewing a release, auditing a subsystem, or before tagging.

guardana/guardana · 53 tokens

cut-a-release

Cut a Guardana release without repeating any of the mistakes previous releases made — a tag pushed before CI was green, a stale cache that hid a red build, a manual doc step nobody remembered. Use when asked to release, tag, publish or bump a version.

guardana/guardana · 57 tokens

harness

End-to-end workflow orchestrator. Walks the 11-phase pipeline, invoking each phase skill in order inside an internal loop, yielding at consent gates (/approve-direction, /approve-swarm, /grant-commit), and exiting cleanly on yield/failure/done. Decides swarm-vs-solo at Phase 6. Auto-loops /tdd on integrate failures…

friedbotstudio/baseline · 110 tokens

triage

Triage an incoming request — pick the workflow entry phase (intake / spec / tdd / chore) and record the workflow statefile that the Track Guard reads.

friedbotstudio/baseline · 37 tokens

claude-automation-recommender

Analyze a codebase and recommend Claude Code automations (hooks, subagents, skills, plugins, MCP servers). Use when user asks for automation recommendations, wants to optimize their Claude Code setup, mentions improving Claude Code workflows, asks how to first set up Claude Code for a project, or wants to know what…

friedbotstudio/baseline · 77 tokens