Skip to content
Back to skills

Accept

ASecurity

Manually approve the current working tree for this phase, overriding a review loop that will not converge, without leaving the mode.

  • 2 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 19, 2026
developmentgobash

Security analysis

A100/100

Scanned September 19, 2026

npx -y skills add leinardi/adversarial-review-loop --skill accept --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Accept?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Accept
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/leinardi-accept/badge)](https://www.skillsdirectory.com/skills/leinardi-accept)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: accept
description: Manually approve the current working tree for this phase, overriding a review loop that will not converge, without leaving the mode.
argument-hint: "[reason]"
disable-model-invocation: true
user-invocable: true
allowed-tools: "Bash(${CLAUDE_PLUGIN_ROOT}/scripts/arl.sh accept:*)"
---

# Manually accept the current tree

```!
${CLAUDE_PLUGIN_ROOT}/scripts/arl.sh accept --reason-stdin <<'ARL-ARGUMENTS-EOF'
$ARGUMENTS
ARL-ARGUMENTS-EOF
```

## What just happened

The block above is the output of accepting, which ran **before you had a turn**. It grants exactly one thing: the current working tree is added to `approved_trees`, the same record a passing review would have written. It does not advance the phase and does not complete the activation — the next commit still goes through the ordinary review gate, and the tree it commits must still be this exact one for the gate to let it through without a review.

**If it says nothing was accepted, stop.** The activation could not be accepted into for the reason given — read it and follow the alternative it names (usually `set-phases`, `/adversarial-review-loop:resume`, or a fresh `/adversarial-review-loop:implement <plan.md>`).

**If it accepted a `NEEDS_HUMAN` escalation**, the activation is `ACTIVE` again and commits are gated normally from here.

**If it accepted during a `RECONCILE`**, that reconcile still stands — the acceptance did not clear it. The outstanding recovery reset is still required before the mode will complete.

This command exists for one situation: a review loop that keeps raising new findings without converging. It is not a way to skip review as a matter of course — every other phase, and every tree that changes after this one, is reviewed exactly as before.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.