reviewing-regulatory-prior-art-survey
Skill2 starsChanged 3 months agoMITView on GitHub
Use when reviewing an artifact produced by regulatory-prior-art-survey — a regulatory scope map, one angle's search output, one instrument's extract record, or the regulatory register and its report — and deciding whether it can be built on. Judges numbered conditions covering canonical terms the corpus uses, the sector receipt, per-angle verdicts in both directions, verbatim query recording, the recorded zero, cause evidence with observable status, the date separations, authority as a rank and binding force as an orthogonal fact, and the claim-versus-quote boundary that fabricated citations cross. Also judges whether a paywalled clause was paraphrased, whether a merge crossed a non-comparable dimension, and whether a conflict was escalated rather than resolved. Emits exactly one VERDICT approve or revise with findings naming their condition. It does not revise a thin-but-honest result. Keywords: regulatory review, compliance survey review, citation check, survey quality gate.
In this repository77 other files it gives its agents.
- alembicSkill
- atlassian-rest-opsSkill
- authoring-api-referenceSkill
- authoring-api-specSkill
- authoring-architecture-docSkill
- authoring-data-modelSkill
- authoring-design-systemSkill
- authoring-developer-guideSkill
- authoring-feature-specSkill
- authoring-hi-fiSkill
- authoring-prdSkill
- authoring-release-runbookSkill
- authoring-technical-designSkill
- authoring-test-planSkill
- authoring-user-flowsSkill
- authoring-user-guideSkill
- authoring-wireframesSkill
- biomeSkill
- cloudflare-pages-opsSkill
- code-prior-art-surveySkill
- content-template-gatewaySkill
- design-reviewSkill
- external-content-sanitizerSkill
- github-cli-opsSkill
- integrations-prior-art-surveySkill
- jenkins-rest-opsSkill
- market-competitive-prior-art-surveySkill
- ml-prior-art-surveySkill
- motion-reactSkill
- netlify-opsSkill
- openapi-ts-clientSkill
- platform-ecosystem-prior-art-surveySkill
- polyglot-git-hooksSkill
- project-document-discoverySkill
- pydantic-v2Skill
- python-monorepo-architectureSkill
- react-component-testingSkill
- regulatory-prior-art-surveySkill
- rest-api-designSkill
- reviewing-api-referenceSkill
- reviewing-api-specSkill
- reviewing-architecture-docSkill
- reviewing-code-prior-art-surveySkill
- reviewing-data-modelSkill
- reviewing-design-systemSkill
- reviewing-developer-guideSkill
- reviewing-document-discoverySkill
- reviewing-document-setSkill
- reviewing-feature-specSkill
- reviewing-hi-fiSkill
- reviewing-integrations-prior-art-surveySkill
- reviewing-market-competitive-prior-art-surveySkill
- reviewing-ml-prior-art-surveySkill
- reviewing-platform-ecosystem-prior-art-surveySkill
- reviewing-prdSkill
- reviewing-release-runbookSkill
- reviewing-scale-prior-art-surveySkill
- reviewing-security-prior-art-surveySkill
- reviewing-technical-designSkill
- reviewing-test-planSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 34
So never report what the validator already checks: If your finding could have been produced by the script, it is not a review finding — and …
- Distinctive: line 124
If a finding's remedy lies OUTSIDE the file this artifact's author was asked to write — the map, a registry row, another angle's output — …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Reviewing a regulatory prior-art artifact
You are the judgment half of a two-part gate. The deterministic half has already run: shape, enums,
ranges, arithmetic and reconciliation are checked by validate_regulatory_prior_art.py, and an
artifact reaching you has passed it at exit 0.
So never report what the validator already checks. If your finding could have been produced by the script, it is not a review finding — and repeating it costs a revise round on work that was correct.
What you judge
references/conditions.md is the single source of the bar. Read it and judge against it. The
producer skill assigns every duty; these conditions say how each is JUDGED and add none of their
own.
The one failure worth more attention than the rest
A fabricated citation — an obligation the instrument's text does not carry, an identifier
nobody resolved, or a text_retrievable that is not the state the fetch actually reached. The
third is the reachable route to the same lie: the gate refuses a quote on an unreadable record
before you ever see it, so a fabrication that survives to you is one wearing a state it did not
earn. C19 and C23 are where these surface, and a finding under either is worth more than three
findings under anything else.
Your evidence, and where it lives
| evidence | where |
| the artifact | handed to you in the task |
| the regulatory scope map | handed to you alongside a search output — a separate input, not part of the artifact |
| the scope and classification the producer was handed | handed to you alongside the map — see below |
| the schemas | regulatory-prior-art-survey/schemas/ |
| the source registry | regulatory-prior-art-survey/references/source-registry.yaml |
| the angle reference | regulatory-prior-art-survey/references/angles/<angle_id>.md |
The scope and classification are handed to you because without them meta.classification is
unfalsifiable. The producer is ordered to transcribe them VERBATIM, precisely so a verdict citing
a value can be checked against what it actually had. Judging every verdict against the producer's
own transcription checks that it is self-consistent and nothing more — a fabricated value reads
exactly like a real one. C5 and C8a both rest on this.
Three of the six are PRODUCER-package paths you read; the other three are handed to you. C2 needs the map — its whole test is a
candidate's evidence against the group its found_by names. C15 and C23 need the angle reference
and the registry, and deriving an angle's source list from the artifact is circular: an artifact
that omits a source omits it from any list you read off it.
If you cannot reach what a condition names, say so and emit no verdict on that condition rather than downgrading it silently.
Anything you cannot ground in those six is an OBSERVATION, not a finding — say it as one, plainly, and do not attach a condition to it. An ungrounded finding costs a revise round on correct work — and the number of revise rounds one artifact gets is capped by whatever dispatched you, so a round spent on an ungrounded finding is a round the real ones do not get.
Read outcome first
It decides which conditions apply at all.
outcome | what is owed | what the coverage conditions expect |
ran | cells, and candidates if anything was admitted | every owed (group × source) pair has a cell |
not_run | NOTHING but the verdict it honours | no cells and no candidates is CORRECT |
vacated | cells, their causes, a vacated.cause and a retrieval_summary; no candidates and no unadmitted rows | cells as for ran; an empty candidate list is not a gap |
A not_run artifact has no cell for any pair, and the deterministic gate REQUIRES that. Reading a
coverage condition against it would revise work the other half of the gate certified.
Proportionality
A thin result is not a failed result. A lightly-regulated scope yields few instruments, and a survey that says so honestly is complete. Revise only on a named gap against a numbered condition — never because the artifact could have been longer.
This matters unusually here. It is always possible to name one more instrument that might bind, and a reviewer who does that pushes the producer toward padding a register with obligations nobody in scope has — which is worse than the thinness, because an architect acts on it.
Output
Findings, each naming its condition, then exactly one terminal line:
VERDICT: approveor
VERDICT: reviseNothing after it.
Upstream remedies
If a finding's remedy lies OUTSIDE the file this artifact's author was asked to write — the map, a registry row, another angle's output — they cannot perform it, and ordering the revision anyway burns cycles and ends somewhere naming neither the artifact nor the inconsistency.
Label it UPSTREAM: and name the exact file and field that must change. One that names no
target is a finding you could not localise, and it belongs in Observations.
A file whose every finding is UPSTREAM gets approve. You judge THIS artifact against its own
contract; if it satisfies that contract and the defect is elsewhere, it is not the defective thing.
Reserve revise for a remedy the author can actually perform. There is no third verdict —
escalation is for whoever dispatched you, not for you.
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.

