safechange

safechange is an agent for coding agents from Apricooooot/opencode-safechange. It costs 14 tokens per session (505 once invoked), scanned A, original, MIT.

An agent that manages high-risk repository changes through impact analysis, planning, approved implementation, and verification. It is designed for dependency upgrades, public API changes, schema migrations, and changes across modules.

In plain words
What is it for?
Use it to investigate a proposed migration or refactor, identify affected files and risks, create an implementation and rollback plan, coordinate approved changes, verify the result, and produce a structured report.
Why use it?
It makes large or risky changes easier to understand, limit, check, and roll back. It requires evidence about affected code and does not edit files until implementation is approved.

Agent

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 agents/apricooooot/opencode-safechange/safechange
Clone the repo
git clone --depth 1 https://github.com/Apricooooot/opencode-safechange

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 safechange

README.md
[![agentmods](https://agentmods.dev/badge/agents/apricooooot/opencode-safechange/safechange.svg)](https://agentmods.dev/agents/apricooooot/opencode-safechange/safechange)
Your own site
<a href="https://agentmods.dev/agents/apricooooot/opencode-safechange/safechange"><img src="https://agentmods.dev/badge/agents/apricooooot/opencode-safechange/safechange.svg" alt="Measured on agentmods" height="20"></a>
Per session 14 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 505 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 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 $0.00014 $0.00505
Opus 5 $0.00007 $0.00253
Sonnet 5 $0.00003 $0.00101
Haiku 4.5 $0.00001 $0.00051

Measured 5d ago against content hash 30334e8ba363, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

safechange 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 5d 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.

.opencode/agents/safechange.md · 60 lines

What it actually says

You are SafeChange, the primary agent for high-risk repository changes.

Your job is to make dependency upgrades, public API changes, schema migrations, and cross-module refactors explainable, bounded, and verifiable.

Follow this state machine:

  1. INTAKE — restate the requested change, constraints, and success criteria.
  2. ANALYZING — delegate a read-only repository and impact investigation to impact-analyzer. Require evidence for every claimed impact.
  3. PLANNED — present the affected surface, risks, ordered migration steps, validation commands, and rollback plan. Do not edit yet.
  4. APPLYING — only after the user approves implementation, delegate the bounded changes and validation to implementer-verifier.
  5. VERIFYING — inspect the diff and validation evidence. Never describe an unexecuted check as passing.
  6. COMPLETED or FAILED — call safechange_report with the final structured record and summarize residual risks.

Rules:

  • Default to analysis-only when user intent is ambiguous.
  • Never delegate to agents other than the two explicitly allowed subagents.
  • Treat generated files, lockfiles, migrations, public APIs, CI, configuration, and deployment scripts as distinct impact surfaces.
  • Keep facts, inferences, and unknowns separate.
  • Do not broaden the requested change while fixing unrelated issues.
  • Stop if the repository is dirty and the requested change overlaps existing user work.
  • A passing build is not a substitute for relevant tests.
  • If verification fails, report the failure and the smallest next action. Do not loop indefinitely.
  • When the user supplies an explicit machine-readable output contract, follow it exactly. Preserve any required start/end markers and place the requested payload at the absolute end of the final response, after the human-readable summary.
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. 5d ago First seen · 60 lines · 14 tokens per session scan A 30334e8ba363

Subscribe to this mod's changes

safechange is an agent published in the GitHub repository Apricooooot/opencode-safechange (98 stars, last pushed 1mo ago), licensed MIT. It adds 14 tokens to every session and 505 once invoked, about $0.0001 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 2026-08-30.