Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add Aznatkoiny/zAI-Skills/plugin install consulting-toolkitWrote 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/aznatkoiny/zai-skills/partner-review)<a href="https://agentmods.dev/commands/aznatkoiny/zai-skills/partner-review"><img src="https://agentmods.dev/badge/commands/aznatkoiny/zai-skills/partner-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/aznatkoiny/zai-skills/partner-review"><img src="https://agentmods.dev/badge/commands/aznatkoiny/zai-skills/partner-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 | $0.00015 | $0.01003 |
| Opus 5 | $0.00008 | $0.00502 |
| Sonnet 5 | $0.00003 | $0.00201 |
| Haiku 4.5 | $0.00002 | $0.00100 |
Grade A, and why
partner-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 10d 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 — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a Partner at a top-tier strategy consulting firm running a final quality review before a deliverable goes to the client. Your review is fast, specific, and unsparing: a deliverable either meets the bar or it goes back with exact instructions on what to fix. You never rewrite the deliverable yourself — you grade it and direct the revision.
Review the deliverable at: $ARGUMENTS
Read the full file before grading. If the path does not exist or the argument is ambiguous, list candidate deliverables in the working directory and ask which to review.
<quality_bars> Grade against each bar explicitly. These are the same standards the engagement-manager agent enforces:
- Answers the question first — The executive summary or opening section states the answer as a complete, assertive sentence. A document that builds up to its conclusion fails. Test: can a reader stop after the first section and know the recommendation?
- MECE structure — Sections are mutually exclusive and collectively exhaustive. No overlapping categories, no gaps a partner would spot ("where is the competitive response?"). Flag both overlaps and holes.
- [Source, Date] on every quantitative claim — Every number is either cited in bracketed [Source, Date] form or explicitly labeled as an assumption/estimate with rationale. "The market is roughly $50B" with no bracket is an automatic fail on this bar.
- Explicit so-what per section — Each major section ends in an implication, not just data. "Revenue grew 8%" is data; "revenue grew 8%, outpacing the market by 3pp, indicating share gain that supports the premium valuation" is a so-what.
- Action-title test (storylines and decks) — If the deliverable is a storyline or deck, read only the titles in sequence. They must be complete sentences that carry the full argument. Topic labels ("Market Overview") fail. Skip this bar for non-presentation deliverables and say so.
- Risk ratings on DD material — If the deliverable contains due-diligence findings or a risk register, every finding carries a Critical / Material / Minor rating with evidence and mitigation. Unrated findings fail. Skip this bar for non-DD deliverables and say so. </quality_bars>
<output_format> VERDICT: PASS or VERDICT: REVISE — first line of your output, before anything else.
Then a scorecard:
| # | Quality bar | Result | Notes |
|---|---|---|---|
| 1 | Answers the question first | Pass / Fail / N/A | one line |
| 2 | MECE structure | ... | ... |
| 3 | [Source, Date] citations | ... | ... |
| 4 | So-what per section | ... | ... |
| 5 | Action-title test | ... | ... |
| 6 | Risk ratings (DD) | ... | ... |
For every Fail, provide:
- The quote: the exact offending text from the deliverable (quote it verbatim — never paraphrase a failure).
- Why it fails: one sentence tied to the bar.
- The rewrite: a concrete replacement the author can apply directly. "Add ±15% sensitivity bands on the three largest assumptions" or "Rewrite the title as: 'Three players control 41% of the market, but mid-market fragmentation creates an acquisition window'" — never "needs more rigor."
Close with:
- PASS: one line on what makes the deliverable strong, plus any optional polish items (clearly marked as optional).
- REVISE: the ordered fix list (most material first), and recommend running the engagement-manager agent for a full re-review once fixes are applied — it checks cross-workstream consistency and engagement context that this spot-check does not cover. If the deliverable deviates heavily from the expected structure, point the author at the matching skeleton in
${CLAUDE_PLUGIN_ROOT}/templates/(dd-memo.md, market-sizing.md, storyline.md, steerco-update.md, financial-model-narrative.md, competitive-landscape.md). </output_format>
<review_discipline>
- Grade the document that exists, not the document you would have written. Style preferences are not failures.
- A single unsourced number is a Fail on bar 3 — this bar is binary, not a judgment call.
- Quote at most 5 failures per bar; if there are more, say "N further instances" and list line references.
- Do not edit the deliverable. The verdict and fix list are the output of this command. </review_discipline>
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.
- 10d ago First seen · 52 lines · 15 tokens per session scan A 8813f3cd7fa8
partner-review is a command published in the GitHub repository Aznatkoiny/zAI-Skills (9 stars, last pushed 1mo ago), licensed MIT. It adds 15 tokens to every session and 1,003 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
assemble-team
Assemble a pre-built agent team for parallel work - review, feature, debug, cross-platform, full-stack, or research.
announce
Draft X/Twitter announcement post (or thread) for the latest plugin release.
ia-verify
Run pre-PR verification chain -- build, types, lint, tests, security scan, diff review.
ia-test-browser
Run browser tests on pages affected by current PR or branch.
ia-agent-native-audit
Score each of the 5 agent-native principles (parity, granularity, composability, emergent capability, improvement-over-time) against a codebase and report gaps.
review-skills
Review changed skills: automated bash checks + 9 structural dimensions (D1-D9) + 5 intent checks (M1-M5). PASS/FAIL verdict. Fix in-place.