deviation

A safety instruction set for situations where the agent is unsure of the user's intent, encounters an unexpected result, or cannot reliably complete the task.

In plain words
What is it for?
Use it to pause changes, recheck the request, challenge the current approach, escalate for review, and record the cause of a deviation.
Why use it?
It prevents the agent from continuing based on risky assumptions when an important detail is unclear or something has gone wrong.

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/griddynamics/rosetta/deviation
Any agent
npx skills add griddynamics/rosetta --skill deviation
Clone the repo
git clone --depth 1 https://github.com/griddynamics/rosetta

Made for: Claude Code, Codex.

Per session 81 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 269 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 $0.00081 $0.00269
Opus 5 $0.00041 $0.00134
Sonnet 5 $0.00016 $0.00054
Haiku 4.5 $0.00008 $0.00027

Measured 2d ago against content hash 7076b3bf0a43, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

deviation 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 2d 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.

instructions/r2/core/skills/deviation/SKILL.md · 33 lines

What it actually says

  1. STOP all changes immediately.
  2. DOUBLE CHECK against original intent.
  3. "THINK THE OPPOSITE" — challenge current direction.
  4. Escalate: subagents → orchestrator → user.
  5. State briefly: understood, conflicted, unresolvable.
  6. Wait for explicit decision.
  7. Update AGENT MEMORY.md with root cause.
  8. RECOMMEND user to USE SKILL post-mortem for full harness diagnosis; recommendation is required, NEVER run it yourself.
  • Rationalizing deviation as "improvement".
  • Continuing "just a bit more" after detection.
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. 2d ago First seen · 33 lines · 81 tokens per session scan A 7076b3bf0a43

Subscribe to this mod's changes

deviation is a skill published in the GitHub repository griddynamics/rosetta (341 stars, last pushed 4d ago), licensed Apache-2.0. It adds 81 tokens to every session and 269 once invoked, about $0.0004 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.

Related

Other skills, from other repositories

unity-mcp-orchestrator

Orchestrate Unity Editor via MCP (Model Context Protocol) tools and resources. Use when working with Unity projects through MCP for Unity - creating/modifying GameObjects, editing scripts, managing scenes, running tests, or any Unity Editor automation. Provides best practices, tool schemas, and workflow patterns for…

CoplayDev/unity-mcp · 72 tokens

fetch-url

Fetch a single URL and convert its content to Markdown. Use when you need to read a web page, documentation page, or API reference without indexing it. The content is returned as plain Markdown text on stdout.

arabold/docs-mcp-server · 45 tokens

signing-entitlements

Inspect signing, entitlements, hardened runtime, and Gatekeeper issues for macOS apps. Use when asked to diagnose code signing failures, missing entitlements, sandbox problems, notarization prerequisites, or trust-policy launch errors.

robinebers/openusage · 49 tokens

macos-swiftpm

Build, run, and test SwiftPM macOS packages and executables. Use when the repo is package-first or has no Xcode project.

robinebers/openusage · 35 tokens

macos-test-triage

Triage macOS tests across Xcode and SwiftPM. Use when narrowing failures, explaining assertions or crashes, or separating setup from regressions.

robinebers/openusage · 36 tokens

fix-codesign-error

Slash command that inspects a macOS signing or entitlement failure and explains the minimum fix path. Invoke explicitly with /fix-codesign-error — this skill never self-triggers.

robinebers/openusage · 42 tokens