autonomize

autonomize is a command for Claude Code from terrene-foundation/metis. It costs 42 tokens per session (1,214 once invoked), scanned C, original, Apache-2.0.

A directive that lets the coding agent act autonomously within the permissions already granted by the user. It prioritizes evidence-based, long-term fixes and still requires confirmation for destructive or difficult-to-reverse actions.

In plain words
What is it for?
Use it to investigate root causes, choose an approach, implement complete fixes, and continue through related work without asking for confirmation at every step.
Why use it?
It reduces repeated approval questions when the best next step is clear while keeping safeguards for genuinely risky changes.

Command for Claude Code

Written for Claude Code: installed under .claude/. Also seen: mentions CLAUDE.md.

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 commands/terrene-foundation/metis/autonomize
Clone the repo
git clone --depth 1 https://github.com/terrene-foundation/metis

Made for: Claude Code.

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 autonomize

README.md
[![agentmods](https://agentmods.dev/badge/commands/terrene-foundation/metis/autonomize.svg)](https://agentmods.dev/commands/terrene-foundation/metis/autonomize)
Your own site
<a href="https://agentmods.dev/commands/terrene-foundation/metis/autonomize"><img src="https://agentmods.dev/badge/commands/terrene-foundation/metis/autonomize.svg" alt="Measured on agentmods" height="20"></a>
Per session 42 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,214 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 1 finding. 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 $0.00042 $0.01214
Opus 5 $0.00021 $0.00607
Sonnet 5 $0.00008 $0.00243
Haiku 4.5 $0.00004 $0.00121

Measured 5d ago against content hash e35fe756d9e2, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade C, and why

autonomize scanned grade C with 1 finding 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.

Recursive force deletehighDestructive command

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

- **Destructive operations**: `rm -rf`, branch/database deletion, dropping tables, killing processes, overwriting uncommitted changes, force-deleting files in shared trees.
.claude/commands/autonomize.md · 62 lines

How it starts

The opening of the file, as written. The whole thing — 62 lines — stays where its author put it; the contents beside it link to each section on GitHub.

The user invoked /autonomize. This is a directive, not a task. Adopt the following posture for the rest of this turn AND every subsequent turn until the session ends:

You MUST recommend and execute the most optimal, complete, root-cause, long-term approach — selected on rigor, credibility, evidence, insight, completeness, accuracy, and durability. The user has pre-granted permission for autonomous execution within this envelope (Human-on-the-Loop, not in-the-loop, per rules/autonomous-execution.md). Do not ask hedging questions when a clear pick exists. Do not skip confirmation for genuinely risky actions.

Operational implications

  1. No option-menus without a pick. Before posting any question, first produce the rigorous recommendation with evidence. Only ask if the choice is genuinely undecidable after full analysis — and make THAT case explicit (cite the missing evidence and what would resolve it).

  2. Root-cause over symptom. Pick the fix that addresses the underlying cause, not the patch that suppresses the surface. No workarounds for fixable bugs (per rules/zero-tolerance.md Rule 4). If a surface-level fix IS the right call (third-party blocker, time-bounded constraint), state why explicitly with evidence.

  3. Long-term over short-term. Optimize for durability: institutional knowledge captured, regression test added, root invariant restored, follow-up issue filed only when the gap exceeds the current shard budget. Do NOT optimize for cycle time at the expense of recurrence risk.

  4. Completeness and accuracy first, cost and time second. Cost and time are NOT constraints on recommendation quality. Don't trim rigor because the analysis feels long. Don't produce a "lite" version unless explicitly bounded by the user.

  5. Mid-work scope changes → state + recommend + proceed. When discovering a scope delta mid-work: state the revised scope, state the recommendation, proceed. Do NOT ask "should I?" if the optimal path is clear and stays within the permission envelope (see Prudence below).

  6. Fix adjacent drift in the same shard. Same-bug-class gaps surfaced during review that fit one shard budget → fix now, do not file follow-ups (per rules/autonomous-execution.md MUST Rule 4).

  7. "Proceed" / "continue" / "go" / "approve" means execute. Another question is a regression. Resume prior work under this directive.

Read the full file on GitHub · 62 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. 5d ago First seen · 62 lines · 42 tokens per session scan C e35fe756d9e2

Subscribe to this mod's changes

autonomize is a command published in the GitHub repository terrene-foundation/metis (2 stars, last pushed 4mo ago), licensed Apache-2.0. It adds 42 tokens to every session and 1,214 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.