loop-audit

A command for checking an automated agent loop against five safety rules.

In plain words
What is it for?
Use it to review a loop for a binary exit condition, iteration and spending caps, sandboxing, and approval before pushing, deploying, or sending.
Why use it?
It reveals missing limits, isolation, exit checks, or human approval before risky actions.

Command for Claude Code

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/fltman/loop-engineer/loop-audit
Clone the repo
git clone --depth 1 https://github.com/fltman/loop-engineer

Made for: Claude Code.

Per session 12 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 252 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 $0.00012 $0.00252
Opus 5 $0.00006 $0.00126
Sonnet 5 $0.00002 $0.00050
Haiku 4.5 $0.00001 $0.00025

Measured 2d ago against content hash 76ba441ecc9b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

loop-audit 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.

.claude/commands/loop-audit.md · 23 lines

What it actually says

Audit a loop against the loop-engineer five-point guardrail checklist.

Target: $ARGUMENTS

Steps:

  • If a path is given, read it. If only a description is given, work from that (and ask for the script if you need it).
  • Read .claude/skills/loop-engineer/reference/guardrails.md for the standard.
  • Report each of the five guardrails as PASS / FAIL, citing the specific line or the concrete gap:
    1. Verifiable, binary exit condition — checked each iteration
    2. Max-iteration cap — a hard bound in code
    3. Budget cap — summed cost, enforced before the next call
    4. Sandbox — branch/worktree/container, never unattended on main
    5. Human checkpoint — for any push/deploy/send/irreversible action
  • For every FAIL, give a concrete fix (a diff or the exact lines to add).
  • End with a one-line verdict: is this loop safe to run unattended?
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 · 23 lines · 12 tokens per session scan A 76ba441ecc9b

Subscribe to this mod's changes

loop-audit is a command published in the GitHub repository fltman/loop-engineer (35 stars, last pushed 2mo ago), licensed MIT. It adds 12 tokens to every session and 252 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.