models

A read-only command that reports which model a coding-agent run is actually using and the evidence behind that routing choice. Model routing is the process of selecting a model for a run.

In plain words
What is it for?
Use it to inspect a run in text or JSON, including its selected and actual model, effort setting, evidence, fallbacks, and review independence.
Why use it?
It makes model selection and any fallback or degradation visible without changing files or starting a run.

Command

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/lookatitude/guild/models
Clone the repo
git clone --depth 1 https://github.com/lookatitude/guild
Per session 115 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,258 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 $0.00115 $0.01258
Opus 5 $0.00057 $0.00629
Sonnet 5 $0.00023 $0.00252
Haiku 4.5 $0.00012 $0.00126

Measured yesterday against content hash d2e5e1cc3025, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

models 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 yesterday.

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.

commands/models.md · 96 lines

How it starts

The opening of the file, as written. The whole thing — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/guild:models — model-routing inspection (read-only)

One sub-verb: inspect. It answers "which model is this run actually using, on what evidence?" without changing anything. No phase — it acts on a run that already exists. Read-only R: it writes no file, creates no directory, and starts no run.

Args & local flags

  • Args: inspect (required; the only public sub-verb — adding another is a requires-confirmation change)
  • --run-id <id> — inspect this run. Without it the command uses the intake-only candidate run id and labels that in the output, so you can see which run you are reading.
  • --json — emit the same view model as JSON. Text and JSON render from ONE view model, so no host surface can show different labels than another.
  • --cwd <repo-root> — the repo that owns .guild/ (default: current dir).

Gates

None — R (read-only). No run-start preflight and no run recording: this command deliberately does not open, touch, or close a run.

Output

Prints the report (no file written).

Dispatch

Run the command CLI and show its output verbatim:

npx tsx ${GUILD_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT:-$HOME/.local/share/guild/dist/claude-code}}/scripts/models-cmd.ts inspect \
  --cwd "$(pwd)"
# add --run-id <id> to pin a run; add --json for the machine view

Exit codes: 0 report emitted · 1 usage error · 2 no run could be resolved (pass --run-id) · 3 failed closed — either the per-field display validator refused a field (it printed <REJECTED:<category>:<12-hex>> in its place, in text and in JSON) or the fully rendered report tripped the canonical redaction applier. On 3, report the failure and the field name; never reconstruct the rejected value, and never re-read the artifact to "show what it really said" — the refusal is the point.

Every row is REBUILT, never echoed. A persisted inspection report under .guild/runs/<id>/inspection/ is a plain writable file, so it is treated as an UNTRUSTED POINTER: it supplies the label only, and every claim is rebuilt through the canonical inspection service from artifacts that verify themselves (the run's frozen guild.session_context.v1, the content-addressed catalog snapshot). If the persisted file asserts something those artifacts do not support, the output prints UNTRUSTED CLAIM(S) NOT SUPPORTED BY VERIFIED ARTIFACTS and NAMES the fields — present that line verbatim; it means someone or something wrote a claim the evidence does not back.

Read the full file on GitHub · 96 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. yesterday First seen · 96 lines · 115 tokens per session scan A d2e5e1cc3025

Subscribe to this mod's changes

models is a command published in the GitHub repository lookatitude/guild (7 stars, last pushed 2d ago), licensed MIT. It adds 115 tokens to every session and 1,258 once invoked, about $0.0006 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 2026-08-31.