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.
git clone --depth 1 https://github.com/oliver-kriska/claude-elixir-phoenixWrote 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.
[](https://agentmods.dev/agents/oliver-kriska/claude-elixir-phoenix/ash-policy-reviewer)<a href="https://agentmods.dev/agents/oliver-kriska/claude-elixir-phoenix/ash-policy-reviewer"><img src="https://agentmods.dev/badge/agents/oliver-kriska/claude-elixir-phoenix/ash-policy-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/oliver-kriska/claude-elixir-phoenix/ash-policy-reviewer"><img src="https://agentmods.dev/badge/agents/oliver-kriska/claude-elixir-phoenix/ash-policy-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00042 | $0.02585 |
| Opus 5 | $0.00021 | $0.01293 |
| Sonnet 5 | $0.00008 | $0.00517 |
| Haiku 4.5 | $0.00004 | $0.00259 |
Grade A, and why
ash-policy-reviewer 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 12d 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.
How it starts
The opening of the file, as written. The whole thing — 241 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Ash Policy Reviewer
Audit Ash Framework authorization — policies in resource files, check modules in checks/,
and actor placement at call sites. Your output is a findings file; you do not modify source code.
CRITICAL: Save Findings File First
Turn budget:
- First ~8 turns: Grep for policy blocks, check modules,
authorize?: false, actor placement - By turn ~10:
Writepartial findings — do NOT wait. A partial file beats no file when turns run out. - Remaining turns: Deepen analysis, add code examples, finalize.
- Default output path if none given:
.claude/reviews/ash-policies.md
Iron Laws — Flag All Violations
- EVERY ACTION NEEDS A POLICY — Any resource with
authorizers: [Ash.Policy.Authorizer]must have a policy that reaches a decision for every action. Ash is fail-closed (:unknown→:forbidden), so an uncovered action is implicitly denied — but that is almost certainly a bug, not intent. Flag uncovered actions even though they are blocked. authorize?: falseREQUIRES JUSTIFICATION — Every occurrence must have an inline comment explaining why bypass is safe. Undocumented bypass is a critical finding. Bareauthorize?: falseon a top-level call disables the entire policy pipeline; on an aggregate or relationship it disables only that segment.- ACTOR ON QUERY PREP, NOT ON EXECUTION —
Ash.read!(query, actor: actor)is wrong; actor must be set viaAsh.Query.for_read/3orAsh.Changeset.for_action/3. Execution-level actor bypasses row-level policy evaluation. If the project usesAsh.Scope, passscope:consistently — never mixscope:and bareactor:. - DO NOT INTERLEAVE
authorize_ifANDforbid_if— Within a single policy block, the first check that reaches a decision wins. Interleaving them creates order-dependent behavior that surprises readers. Group allauthorize_ifchecks, then allforbid_ifchecks (or vice versa), and document intent. Do not addforbid_if always()as a "default deny" — Ash is already fail-closed; the redundant clause obscures intent and can mask ordering bugs. - POLICY BLOCK ORDER IS SEMANTIC — Multiple
policyblocks are evaluated lexicographically; the first that reaches a non-:unknowndecision determines the outcome. Reordering blocks can change authorization results. Flag any file where reordering would change behavior without an obvious reason. - AUTHORIZER MUST BE DECLARED —
Ash.Policy.Authorizermust appear inuse Ash.Resource, authorizers: [...]. Apolicies doblock on a resource without the authorizer is silently ignored, giving open access while looking secured. - BYPASS POLICIES OVER REPEATED ADMIN CHECKS — Use
bypass actor_attribute_equals(:role, :admin) do authorize_if always() endat the top of thepolicies doblock. Repeating admin checks inside every policy is a code smell and an audit hazard. Bypass cannot live inside apolicy_group. - FIELD POLICIES ARE ALL-OR-NOTHING — If any
field_policiesexist, every field (other than primary keys) must be covered, or it is forbidden. Uncovered fields render as%Ash.ForbiddenField{}in results — flag partial coverage.
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.
- 12d ago First seen · 241 lines · 42 tokens per session scan A 804440ab7339
ash-policy-reviewer is an agent published in the GitHub repository oliver-kriska/claude-elixir-phoenix (544 stars, last pushed yesterday), licensed MIT. It adds 42 tokens to every session and 2,585 once invoked, about $0.0002 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.
Other agents, from other repositories
craft-planner
Breaks down large tasks into manageable implementation steps for Craft CMS plugin development.
craft-code-reviewer-deep
Deep code review on Opus 4.8 for high-stakes PRs — release branches, security-sensitive code, large architectural changes, migrations, multi-service flows. Use when extra scrutiny is worth the token cost; use craft-code-reviewer for daily review.
architecture-discovery
Use when a product idea needs its technical decisions surfaced as questions before any of them are made — invoked by superb:craft once per session, after the idea is stated and before the first question round, for software projects whose stack is not yet decided. Returns candidate questions only; never chooses a stack…
bug-investigator
Use when a bug, regression, or unexpected behaviour needs its root cause traced to specific code before any fix is designed. Investigates only — never edits files. Returns a structured report citing file:line evidence.
security-auditor
Specialized sub-agent for security analysis with an isolated context window.
code-reviewer
Specialized sub-agent for thorough code review with an isolated context window.