adl: Command for Claude Code

.claude/commands/goal-verify.md

goal-verify is a command for Claude Code from frontier-infra/adl. It costs 13 tokens per session (527 once invoked), scanned C, original, MIT.

A command that checks a goal manifest—a file listing the required checks for a piece of work—and records the results as proof.

In plain words
What is it for?
Use it to verify a worker's code or documentation slice, record check results, inspect the actual diff, and approve or reject completion.
Why use it?
It prevents work from being marked complete without running every required check. It also catches changes outside the allowed files or changes that match forbidden patterns.

Command for Claude Code

Written for Claude Code: argument-hint in frontmatter. Also seen: reads .claude/ paths.

This is frontier-infra/adl's own configuration. It tells Claude Code how to work on adl itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything adl configures →

Reuse

Borrowing it

Nothing to install: this file belongs to frontier-infra/adl. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/frontier-infra/adl/main/.claude/commands/goal-verify.md
Clone the repo
git clone --depth 1 https://github.com/frontier-infra/adl

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 goal-verify

README.md
[![agentmods](https://agentmods.dev/badge/commands/frontier-infra/adl/goal-verify.svg)](https://agentmods.dev/commands/frontier-infra/adl/goal-verify)
Your own site
<a href="https://agentmods.dev/commands/frontier-infra/adl/goal-verify"><img src="https://agentmods.dev/badge/commands/frontier-infra/adl/goal-verify.svg" alt="Measured on agentmods" height="20"></a>
Per session 13 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 527 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 1 finding. A grade says what 26 rules found in the file — not that it is safe.
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 $0.00013 $0.00527
Opus 5 $0.00006 $0.00264
Sonnet 5 $0.00003 $0.00105
Haiku 4.5 $0.00001 $0.00053

Measured 7d ago against content hash be7e97046ac0, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade C, and why

goal-verify 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 7d 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.

- Check command unsafe in a verification context (e.g. `rm -rf`, network mutations) → refuse, surface to operator
.claude/commands/goal-verify.md · 38 lines

How it starts

The opening of the file, as written. The whole thing — 38 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Verify a /goal manifest by running every declared check and writing the proof artifact. This is the only way a worker's slice can be marked done — the Stop hook reads the proof and refuses to unblock the worker until signed_off: true.

What to do

  1. Adopt the Warden role. Read .claude/agents/WARDEN.md and apply its discipline strictly. You are not the worker. You do not patch failures. You do not edit the manifest.

  2. Load the manifest at .claude/goals/$ARGUMENTS.yaml. If the file does not exist, refuse — there is nothing to verify against. Compute its SHA-256 and record it in the proof.

  3. Run each check in declaration order. Do not short-circuit on failure. Capture exit codes, stdout, stderr, and duration for each.

  4. Compute the diff. Diff against manifest.base_ref (default HEAD's parent). Validate touches_only and forbidden globs against the actual changed paths. A path outside touches_only or matching forbidden is a check failure, not a warning.

  5. Apply sign-off rules per WARDEN.md:

    • All machine checks passed, no human-review present → signed_off: true
    • Any machine check failed → signed_off: false, populate kickback_reason
    • All machine checks passed, human-review present → signed_off: "pending", populate kickback_reason with the items awaiting operator
  6. Write the proof to .claude/goals/$ARGUMENTS.proof.json. Overwrite any prior proof for the same task-id (re-verification is allowed; the latest proof wins).

  7. Report to the operator per Warden's handoff protocol: task-id, manifest SHA, per-check results, final signed_off, kickback target role if applicable, proof path.

Refusals

  • Manifest missing or unreadable → refuse, surface to operator
  • Check type not in the schema → refuse, do not invent semantics
  • Check command unsafe in a verification context (e.g. rm -rf, network mutations) → refuse, surface to operator

Argument

The task-id to verify follows. It must match an existing manifest filename (without .yaml).

Read the full file on GitHub · 38 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. 7d ago First seen · 38 lines · 13 tokens per session scan C be7e97046ac0

Subscribe to this mod's changes

goal-verify is a command published in the GitHub repository frontier-infra/adl (2 stars, last pushed 2mo ago), licensed MIT. It adds 13 tokens to every session and 527 once invoked, about $0.0001 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.