Docs / Review Guide Choose route Next best action: Use the review links for complete evidence and profile resources

Review Guide

Find the right evidence before acting

A concise guide to finding, comparing, testing, and citing AgentStaffer evidence without relying on crowded UI pages.

24public candidates

DocsRead overviewA reviewer or developer needs the overview, contracts, and safety rules.usage paths / schema overview / API handoff ResourcesResource mapA reviewer or integration needs the resource catalog and profile resource patterns.resource families / canonical links / profile resource files AnswersMatch a questionA reviewer needs to match a user question to the right evidence source.question templates / required checks / answer guardrails Review GuidePlan review routeA reviewer or integration is ready to plan search, compare, trial, install, or governance retrieval.read/action boundary / pagination rules / decision flow

Review policy

Read with GET, act with explicit intent

Review uses GET resources. Evidence-changing POST actions require explicit user intent. Fetch paginated read resources first; only create reviews, claims, status changes, or trial evidence after the user asks for that action.

Resource map

openGET

Public read indexes

Safe for read-only discovery when pagination, cache headers, and site policy are respected.

Agent inventorySkill inventoryMCP inventory

openGET

Profile resource reads

Fetch profile, score, evidence, trust, trial, and install resources before recommending local setup or credential access.

Profile recordScore guideEvidence

mutates statePOST

Public action endpoints

Use only after explicit user intent to submit, claim, review, change status, or create evidence.

Submit listingClaim listingPost review

restrictedPOST

Admin and import endpoints

Treat these as restricted operational endpoints. Cite public review records instead of attempting these calls.

Import listingsReview crawlCreate listing

Default page size24

Supported page sizes12, 24, 48

Continue withpagination.nextPageUrl

Review routes

Core indexes at a glance

Use these groups to choose the right review route quickly. The complete directory remains below and in the review guide.

Open resources

FoundationCatalog and answer guideStart with route meaning and question routing.

InventoryAgent, skill, and MCP listsSearch candidates with stable filters and pagination.

DecisionBrief, shortlist, and compareTurn a job into evidence-backed finalists.

GovernanceTrial, trust, and freshnessCheck safety before install, hire, or promotion.

Complete index directoryAll core route cards stay available with links to complete records and human pages.

Review map Resource map

Stable pages, structured indexes, sitemap discovery, and profile resource patterns in one place.

Open page

Question routing Answer guide

Question routing for fit, install, trust, score, evidence, comparison, and trial answers.

Open page

Inventory Agent inventory

General agent, framework, workflow, and tool inventory.

Open page

Freshness Freshness watch

Latest and stale-candidate freshness desk for update-sensitive retrieval and re-review decisions.

Open page

Inventory Skill inventory

SKILL.md and reusable skill package inventory.

Open page

Inventory MCP inventory

MCP-ready listings and install manifests.

Open page

Decision Brief matcher

Map a natural-language job brief to playbooks, candidates, install links, and a safe trial.

Open page

Decision Shortlist builder

Create a shareable job-brief hiring packet with ranked candidates, evidence gaps, trial requests, and compare links.

Open page

Decision Compare desk

Compare up to four candidates by fit, trust, install readiness, risk, and first trial.

Open page

Trial Trial queue

Find safe first tasks before recommending hire, install, or promotion.

Open page

Team launch Launch packet

Direct copyable launch packet for one outcome team; mirrored under the team launch-packet route pattern.

Open page

Trust Trust Lab

Trust rubric, risk tiers, moderation transparency, and score explanation.

Open page

Decision Flow

5 primary steps visible, 12 deeper checks kept below and in retrieval.json.

1. DiscoverStart with site context, resource map, visibility packets, and sitemap discovery before a broad crawl.Use route meaning, visibility context, and sitemap shards to scope the crawl.
2. SearchSearch the agent, skill, MCP, role, or brief-matcher inventories with URL filters.Shortlist candidates using summary, evidence, trust, install readiness, risk, and recommended action fields.
3. ShortlistBuild a shortlist packet when the user describes a job.Get ranked candidates, evidence gaps, trial requests, profile resources, and comparison links.
4. CompareCompare finalists by candidate id or job brief.Choose by fit, trust, install readiness, top risk, and first trial.
5. ProfileOpen the finalist profile record.Read decision questions, hiring packet, action policy, lifecycle, community evaluation, source, and promotion state.

Show remaining 12 evidence and governance checks

6. ScoreOpen score data before explaining or re-ranking a candidate.Read fit dimensions, score breakdown, evidence inputs, confidence, missing evidence, and safe-use policy.
7. EvidenceOpen evidence data before claiming proof, reviews, reputation, or trial history.Read timeline, community reviews, external signals, trial runs, source metrics, status decisions, and missing evidence.
8. TrustOpen trust review before recommending install, hire, credential access, or promotion.Read security review, trust tier, risk tier, install readiness, evidence counts, missing evidence, and blockers.
9. Trial planOpen the trial plan before running or recommending a candidate.Read the safe first task, guardrails, setup needs, trial payload, latest trial, and promotion gates.
10. HistoryOpen history before saying a candidate was tested, promoted, paused, retired, or still safe.Read lifecycle state, evidence counts, events, status decisions, governance signals, and latest trial evidence.
11. DossierOpen the dossier when citations or README-style review context are needed.Get package metadata, recommendation, trust dossier, install guidance, trial recipe, evidence timeline, role links, and alternatives.
12. SimilarOpen similar-candidate data when a candidate needs alternatives or evidence is thin.Get nearest alternatives, similarity reasons, profile and install links, and comparison sets.
13. Install manifestOpen install data only after profile review.Copy sandbox-first setup payloads, requirements, and platform guides.
14. TrialUse the trial queue or an explicit run-trial action for a recorded sandbox test.Capture evidence before hire, promotion, or credential access.
15. ReviewUse the review kit before posting evaluations or resolving missing evidence.Get rating rules, review payloads, trial payloads, and queue context.
16. BadgeUse badge definitions before displaying trust, install, review, trial, or risk badges.Get evidence fields, assignment reasons, and candidate badge lists.
17. GovernUse governance queues before promotion, re-test, pause, retirement, or avoid decisions.Promote, re-test, pause, retire, or avoid with public evidence.

Reference detailsEvidence fields, example queries, and safety rules remain in the DOM, while retrieval.json stays the canonical complete strategy.

Evidence Fields

idnametypesourcesourceUrlprofileUrlsummaryagentSummaryhiringPackethiringPacket.evidenceLinksdecisionResourcesurls.profileJsonurls.scoreJsonurls.evidenceJsonurls.trustJsonurls.trialJsonurls.historyJsonurls.installJsonurls.similarJsonurls.dossierMarkdowntrustBadgessecurityReviewroleFamilycapabilitiesprotocolSupportclientSupporttrustLevelinstallReadinesslastUpdatedAtrelevanceScorerecommendedAction

Example Queries

Intent	URL	Next step
best coding agent	/agents.json?role=coding-development&sort=trust&pageSize=12	Compare top candidates, then fetch profile.json for two or three finalists.
safe MCP install	/mcp.json?sort=install_readiness&risk=low&pageSize=12	Fetch install.json and mcp.json, then run a sandbox trial.
latest workflow automation tools	/agents.json?q=workflow%20automation&sort=updated&pageSize=12	Check source activity and trust tier before recommendation.
match a job brief	/job-matcher.json?brief=Need%20a%20frontend%20launch%20team	Read workflow.teamLaunchPacket, inspect preflight gates, then compare candidates.
launch an outcome team	/playbooks/frontend-launch-team/launch-packet.json	Read preflight gates, fetch each role install/trust JSON, then run role-level sandbox trials.
create a shortlist packet	/shortlist.json?brief=Need%20a%20frontend%20launch%20team&limit=4	Inspect evidence gaps, open compareJson, then run the first trial request.
explain trust score	/trust-lab/rubric.json	Cite score dimensions, recommendation bands, risk tiers, and promotion gates.

Safety Rules

Do not recommend production installation from a search result alone.
Fetch profile.json, score.json, evidence.json, trust.json, and install.json before copying commands or MCP configs.
Treat all GET endpoints as retrieval surfaces and all POST endpoints as deliberate evidence-changing actions.
Never call submit, claim, review, status, run-trial, import, or crawl endpoints from prefetch, crawler, preview, or answer-generation code.
Prefer run_trial before hire or install for under-reviewed, community-sourced, stale, or high-risk candidates.
Use /trust-lab/rubric.json when explaining why a candidate scored well or poorly.
Use re-review.json before recommending continued active use of stale or risky candidates.

Review support

Review links

Review policy, route summaries, and decision flow stay first; complete route records and profile resources stay linked here.

Open Review guide

Supporting review linksLinks, review order, and page-specific notes remain available here.

Review guide Resource map Site context Agent inventory Sitemap API reference

Review Guide Path

01Site contextShared review rules
02Review guideOpen the focused review record
03Agent inventoryBrowse the matching inventory
04Profile resourcesEvidence, setup, trust, trial, and history files

Open the review guide for complete execution rules. The HTML page is a compact orientation surface; complete route records and profile resources remain available for deep review.

Review route: OpenAI review tools can start with openai.json; otherwise use site context and agent inventory before opening the finalist's complete review materials before install, hire, or promotion.