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.
npx agentmods add commands/akholod/consensus-review/doctorgit clone --depth 1 https://github.com/akholod/consensus-reviewWrote 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/commands/akholod/consensus-review/doctor)<a href="https://agentmods.dev/commands/akholod/consensus-review/doctor"><img src="https://agentmods.dev/badge/commands/akholod/consensus-review/doctor.svg" alt="Measured on agentmods" 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 | $0.00029 | $0.00659 |
| Opus 5 | $0.00015 | $0.00329 |
| Sonnet 5 | $0.00006 | $0.00132 |
| Haiku 4.5 | $0.00003 | $0.00066 |
Grade A, and why
doctor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run the diagnostic and present its output.
node "${CLAUDE_PLUGIN_ROOT}/scripts/doctor.mjs" [--offline] [--repo <path>] [--json]
Flags map straight through: offline → --offline, repo=<path> → --repo <path>, json → --json. As everywhere in this plugin, a leading -- is also accepted.
What it is for
This review degrades silently, on purpose. An unavailable consultant marks its lane
unavailable and the run continues. If every consultant fails, the floor case builds the report
from the Claude lane alone and does not abort. gh missing falls back to fetching the diff by URL.
Each of those is right at run time — a partial review beats none — but together they mean a
misconfigured setup yields a report that looks complete and is quietly weaker.
So every check here reports what a non-PASS state does to a review, not just that something
is off. That consequence line is the point of the command.
It checks function, not presence
codex --version succeeds with expired auth and with a schema codex would reject — the exact
failure §6 of /consensus-review calls "a config bug to fix, not a clean review". So the live
probe actually runs codex exec read-only against this plugin's real findings schema, with stdin
closed, exactly as a review does. Likewise the claude probe sends the production discovery prompt
and validates the envelope the adapter parses.
Live probes run by default and cost real calls (roughly a minute, two small model calls). Pass
offline to skip them — the rest is instant and free, but stale auth cannot be detected without
them, and the summary says so.
Sections
| Section | Catches |
|---|---|
| plugin | version skew against the latest release; Node below the required 20 |
| consultants | codex missing (every review becomes single-source), pi, gh present and authenticated |
| code graph | codegraph CLI, a graph in the repo, and whether that graph is older than the last commit — agents are told "absent/stale → grep" but cannot detect staleness from inside a review |
| contracts | the five agents resolve and keep their read-only guarantee; generated prompt blocks match the canonical schema; which model each dimension runs on |
| workspace | a git work tree with commits; .reviews/ writable and ignored — an unignored report becomes an untracked change the next review picks up |
| live | the claude -p envelope, and codex accepting our schema |
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.
- 4d ago First seen · 49 lines · 29 tokens per session scan A cb009f6717f9
doctor is a command published in the GitHub repository akholod/consensus-review (3 stars, last pushed 15d ago), licensed MIT. It adds 29 tokens to every session and 659 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 2026-08-31.
Other commands, from other repositories
adversarial-review
Run an adversarial Gemini code review that challenges the implementation approach and design choices.
review
Run a standard Gemini code review of recent git changes.
logic-fix-all
Autonomous audit-and-fix — after consent, scan the target, fix every logic issue found (all severities), verify each fix, and report anything unresolved.
logic-diff
Check two code versions for semantic equivalence — use after a refactor or rewrite.
logic-explain
Trace execution step by step — use when code behavior is surprising or confusing.
logic-health
Sweep a whole codebase or directory for logic correctness — use before a release or to identify risk hotspots.