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/tznthou/claude-prismWrote 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/tznthou/claude-prism/pi-multi-review)<a href="https://agentmods.dev/commands/tznthou/claude-prism/pi-multi-review"><img src="https://agentmods.dev/badge/commands/tznthou/claude-prism/pi-multi-review/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/commands/tznthou/claude-prism/pi-multi-review"><img src="https://agentmods.dev/badge/commands/tznthou/claude-prism/pi-multi-review.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 | $0.00013 | $0.06711 |
| Opus 5.5 | $0.00005 | $0.02684 |
| Sonnet 5.5 | $0.00003 | $0.01342 |
| Haiku 4.5 | $0.00001 | $0.00671 |
Grade A, and why
pi-multi-review 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 27d 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 — 502 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Multi-Provider Adversarial Review (Codex + Gemini + Claude)
Send the same code to both Codex and Gemini for adversarial review with divided attack surfaces, then Claude synthesizes. Codex attacks security & data integrity; Gemini attacks design, UX & maintainability. Three adversarial perspectives, maximum blind spot elimination.
Execution
Dispatch rules (v0.13.0+): For parallel provider consultation, use the
Agenttool withsubagent_type: "general-purpose"to spawn two parallel sub-agents — NOT twoBashtool calls in a single main-conversation response. Main-conversation Bash is a structural FIFO queue: the second Bash waits for the first to finish (delta ≈ first_execprecisely, measured N=7 across two capacity slots), which defeats the parallel intent. Sub-agent fan-out dispatches Bash calls truly in parallel (median delta 2.8s, N=13), saving roughly 28% wall-clock on a two-provider call.Inside each sub-agent, run
call-codex.sh/call-gemini.shin foreground synchronous mode withtimeout: 600000(Bash tool's 10-minute ceiling). Do not use&,nohup, orrun_in_background: true— Claude Code 2026-04+ has an auto-background child-lifecycle regression that silently kills the child (output file stays 0 bytes). If a sub-agent's Bash returns empty stdout, the Bash template below falls back to reading the caller-owned$OUT_PATHfile (v0.14.3+, wrappertees here viaCLAUDE_PRISM_OUT_TMP).
0. Pin the working directory
Before any other step, run pwd and keep the result as <WORKDIR>. Sub-agents start in whatever directory the main conversation is in when it dispatches them; if you cd elsewhere in between (a scratch directory to try something out, say), the wrapper would run outside this project. The template in Step 3c sends each sub-agent back to <WORKDIR> first.
1. Determine review scope
Same as /pi-code-review, based on $ARGUMENTS:
- No args → staged changes
- File path → specified file
--diff→ unstaged changes--pr→ PR diff--verbose→ also show issues that were filtered out (< 80 confidence) with their scores
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.
- 27d ago Changed · +24 lines a0be8f31ffdd
- 2mo ago First seen · 478 lines · 13 tokens per session scan A 347fe20f7a68
pi-multi-review is a command published in the GitHub repository tznthou/claude-prism (10 stars, last pushed yesterday), licensed MIT. It adds 13 tokens to every session and 6,711 once invoked, about $0.0001 per session on Opus 5.5. 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.
pi-multi-review compared with commands from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| pi-multi-reviewtznthou/claude-prism | Command | 10 | yesterday | MIT | A | 13 | Claude Code |
| doctoropensesh/KARIMO | Command | 290 | 25d ago | Apache-2.0 | A | 0 | Claude Code |
| greptile-reviewopensesh/KARIMO | Command | 290 | 25d ago | Apache-2.0 | A | 0 | Claude Code |
| conventions-checkeugenelim/agent-ready-repo | Command | 28 | yesterday | Apache-2.0 | A | 28 | Claude Code |
| register-plugin-assets977043/river-review | Command | 7 | today | MIT | A | 28 | Claude Code |
| logic-healthhyhmrright/logic-lens | Command | 25 | 7d ago | MIT | A | 20 | Claude Code |
| planopensesh/KARIMO | Command | 290 | 25d ago | Apache-2.0 | C | 0 | Claude Code |