adversarial-review
Skill0 starsChanged 32 days agoMITView on GitHub
Hostile second read of a contract, policy, advice letter or adapted precedent before it is signed, sent, published or relied on. A reviewer who did not draft it and assumes it is wrong until proven right checks the law and citations, wrong-jurisdiction remnants, missing mandatory content, enforceability, drafting mechanics and whether the client's position survived, then returns VERDICT PASS or FIX with blocker, major and minor findings and paste-ready fixes. Loops draft, review, fix, re-review until PASS. Use for "adversarial review", "red team this", "check this properly", "would a partner sign off on this", "what would the other side's lawyer attack", or as the last step before anything goes out.
111 other projects use a version of this.
Adopted by
Most recently taken up by
- moiz-73/glassboxtheir own version, first seen Nov 16, 2026
- Alexpechonov/claude-skillstheir own version, first seen Nov 16, 2026
- anton-abyzov/specwithcodetheir own version, first seen Nov 16, 2026
- Axect/skillstheir own version, first seen Nov 16, 2026
- magnusrodseth/dotfilestheir own version, first seen Nov 16, 2026
- lennoxsaint/Skillstheir own version, first seen Nov 15, 2026
Used with
Projects with it also use rethink, root-cause-analysis and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
fabioespindulaSkillrethink
Reframe a plan, idea, or decision before the user commits to it. Use when the user asks to rethink, step back, pressure-test, gut-check, decide, right-size, simplify, think bigger about, or evaluate a concrete plan or decision in product, strategy, technical architecture, career, life, company direction, or personal priorities. This skill improves decision quality before action: clarify the real question, challenge assumptions, compare alternatives, calibrate ambition, classify the operating mode, surface risks, and recommend a next move. Do not use for broad future exploration without a concrete decision; that belongs to a think-big or research-style skill.
Also in 3 projects that use this
MicrosoftSkillroot-cause-analysis
Use when a Regional Retail Manager asks I'm seeing unusual inventory movement at our Northwest stores. Can you help me understand what's happening?
Also in 3 projects that use this
anasahmed07Skillbetter-auth-best-practices
Skill for integrating Better Auth - the comprehensive TypeScript authentication framework.
Also in 4 projects that use this
DockerSkillfix-issue
Fix a single GitHub issue end-to-end: triage, research, write the fix, review, and create a PR. Use when asked to fix an issue: "fix issue 1234", "resolve #500", "create a PR for issue 200".
Also in 3 projects that use this
MicrosoftSkillpull-request
Write concise, focused pull request descriptions for Wassette — at most three sentences on the what and why, with an explicit note on any breaking public-API change and how users adapt. Use when opening or editing a Wassette pull request.
Also in 3 projects that use this
juriwiensSkillcodex-review
Run code review via Codex CLI
Also in 3 projects that use this
- +1 more from microsoft
Other versions111 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 111
- PrefectHQ/fastmcp★ 29ksame name
- jlowin/fastmcp★ 29ksame name
- mengxi-ream/read-frog★ 10ksame name
- epicenter-md/epicenter★ 4.8ksame name
- EpicenterHQ/epicenter★ 4.8ksame name
- basicmachines-co/basic-memory★ 4.6ksame name
- DanMcInerney/architect-loop★ 626same name
- deadlock-mod-manager/deadlock-mod-manager★ 617same name
- boraoztunc/skills★ 473same name
- syahiidkamil/Software-Engineer-AI-Agent-Atlas★ 404same name
101 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository27 other files it gives its agents.
- business-saleSkill
- business-setupSkill
- cite-checkSkill
- contract-reviewSkill
- disputesSkill
- doc-auditSkill
- draft-contractSkill
- employment-checkSkill
- fundraisingSkill
- ip-brandSkill
- jurisdiction-adaptSkill
- law-auSkill
- law-change-impactSkill
- law-nzSkill
- law-ukSkill
- law-usSkill
- lease-reviewSkill
- legal-adviceSkill
- legal-health-checkSkill
- legal-researchSkill
- marketing-complianceSkill
- nda-triageSkill
- privacy-complianceSkill
- redlineSkill
- scenario-testSkill
- sprintlaw-modeSkill
- website-complianceSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 100 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Every finding carries paste-ready wording or says what information is needed.
- Distinctive: line 26
If you cannot start a subagent, do it yourself in a separate pass: finish drafting, then re-read from the first line as the reviewer below, …
- Very specific: line 53
Write it to a file when working on files (review-<round>.md), otherwise in the reply.
“Common” is measured: kept by at least 6 in 10 of the 100 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Adversarial review
The person who wrote a document is the worst person to check it: they read what they meant, not what they wrote. This skill creates a different reader, one who did not draft it, has no stake in it being right, and is looking for the thing that will embarrass the client later. In Sprintlaw's own precedent work, the first hostile round returned FIX on 25 of 26 documents that their drafters believed were finished. That is the point of it.
Non-negotiables
- The reviewer did not draft the document and does not read the drafter's reasoning first.
- Verdict is PASS or FIX only. PASS means no
[B]and no[M]findings. - Every finding carries paste-ready wording or says what information is needed.
- Law the reviewer could not verify is
[unverified], never assumed correct.
Set up the reviewer
Use a fresh context if you can. If the harness can start a subagent, give it only: the document (rendered text with real numbering, plus the file), the brief or instructions, the jurisdiction pack, any saved sources, and this skill. Do not give it the drafter's reasoning. A reviewer who has read the drafter's justifications inherits their blind spots.
If you cannot start a subagent, do it yourself in a separate pass: finish drafting, then re-read from the first line as the reviewer below, writing findings before you look at your own notes.
The reviewer's stance: a senior lawyer in the governing jurisdiction who did not draft this, assumes it is wrong until proven right, and will be blamed if it fails. Specific and concrete. No praise.
Add panel members by stakes (P9 in sprintlaw-mode). Each reads separately and reports separately:
| Panel member | Looks for | Use when |
| Local partner (always) | Wrong law, missing mandatory content, unenforceable clauses, drafting defects | Always |
| Opposing counsel | Ambiguities to exploit, obligations to escape, the claim they would bring | Contracts under negotiation, demand letters, anything adversarial |
| Judge or regulator, first impression | Would this clause survive? Does it look unfair, misleading or oppressive? | Consumer terms, employment documents, restraints, penalties, published policies |
| The client | Does it do what I asked? Can I understand and operate it? | Anything the client will sign or run their business on |
| Proofreader | Mechanics only: run doc-audit | Always for documents |
The checks, in this order
- Law and citations. List every statute, regulation, code, section, regulator, court and case the document mentions or relies on. For each: right name and year, section exists and says what the document implies, in force now (or future commencement handled). Run
cite-checkfor anything doubtful. Build the citation table. - Remnants. Anything from another jurisdiction or an earlier version: statute names, regulators, courts, entity types and identifiers, tax terms (GST, VAT, sales tax), currency, spelling, state or territory references, defined terms from the old version, dangling references to deleted clauses. Run
doc-audit --jurisdiction <target>, then read for the subtle ones. - Completeness. What a local lawyer expects in this document type that is missing. Mandatory statutory content gaps are blockers. Use the jurisdiction pack's area file and the
draft-contractchecklists. - Legal effectiveness. Clauses that would be void, unenforceable or ineffective: contracting out of non-excludable rights, penalties, unreasonable restraints, unfair terms, invalid execution of a deed, defective trial or probation clauses, deductions without consent, security interests not perfected.
- Drafting quality. Cross-references point to the right clause in the rendered numbering; defined terms defined once, used consistently, none orphaned or undefined; numbering and list punctuation; clauses that contradict each other; placeholders; execution blocks that match the parties and the document type; plain English; "shall"; spelling for the jurisdiction.
- Commercial position. Did the drafting carry every point in the brief? Does it still favour the client as intended, within what the other side would sign? Was any important protection dropped?
For contracts that matter, run scenario-test as part of check 6.
Output
Write it to a file when working on files (review-<round>.md), otherwise in the reply.
VERDICT: PASS | FIX
Round: <n> Reviewer: <panel members used> Read: <what was read>
## Citation table
| Reference in document | Status (verified / wrong / unverified) | Correct version | Source checked |
## Findings
[B] <clause as rendered> | <problem> | <exact fix wording or instruction>
[M] ...
[m] ...- [B] Blocker: wrong law, unenforceable where it matters, mandatory content missing, a remnant with legal effect, a brief requirement not carried. Cannot go out.
- [M] Major: a careful partner would not send it as is.
- [m] Minor: polish.
PASS means no blockers and no majors. Every finding carries wording the drafter can paste in. If you cannot give a fix, say what information is needed.
The loop
- Fix round. The drafter applies every [B] and [M]. A finding may be rejected only with a reason grounded in authority or the brief ("reviewers can be wrong too; the statute decides"), recorded next to the finding.
- Re-review. The same reviewer checks each earlier finding (fixed, not fixed, fixed badly), then looks for regressions the fixes introduced, then re-runs the checks on changed clauses.
- Stop at PASS, or after three rounds. Keep the best version seen; if a round made things worse, go back to the earlier version. Report the honest end state: "PASS on round 2" or "FIX after round 3: two majors remain, listed below".
Several documents at once
When reviewing a set (a precedent bank, a suite of policies, every document affected by a law change), finish with a sweep: collect every bad string and pattern found in any review (a repealed term, a wrong section number, an old entity name) and search the whole set for it. Errors recur across documents and individual reviewers miss them.
Reply
Verdict first; then blockers and majors with fixes (applied or still needed); citation table; what was not checked. If the user asked for the fixes to be applied, hand off to redline for tracked changes.
Limits
- A hostile review reduces risk. It does not certify the document. Where the stakes are high (P15), a locally qualified lawyer should still read the final version.
- The reviewer is only as current as its sources. Law the reviewer could not verify stays
[unverified]in the citation table.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

