# Brali > A practical knowledge library for people, developers, researchers, and AI agents. Brali publishes atomic hacks, executable protocols, a Domain/Topic/Method/Lens ontology, evidence states, reviewed Evidence Decisions, research candidates, canonical identities, a versioned static API, optional read-only MCP tooling, and a generated one-skill-per-hack Agent Skills layer. ## Knowledge model v2 - Domain: broad area of practical life or work. - Topic: concrete problem, capability, or outcome. - Method: named structured approach that may cross Topics. - Lens: transferable way of thinking; a Lens is not evidence. - Hack: smallest reusable practical technique. - Protocol: executable sequence or personal experiment. - Evidence state: `reviewed`, `practical`, `pending-review`, or `restricted`. - Research candidate: discovery metadata; it is never evidence by itself. - Evidence Decision: explicit source-review record with supported claim, unsupported/overstated claims, limitations, and editorial outcome. - Canonical identity: `brali::`; labels, URLs, localized terms, and historical IDs are aliases. Legacy Life Area and Growth Zone fields are compatibility data. `topic-pending` means a concrete Topic still needs editorial classification. Research promotion path: `research candidate -> evidence decision -> hack -> protocol -> ontology -> publish`. ## For AI & developers Start at `/for-ai/` and `/api/v1/index.json`. Ontology coverage is at `/ontology/coverage/`. API v1 exposes `openapi.json`, `topics.json`, `protocols.json`, `hacks.json`, `evidence.json`, `evidence-decisions.json`, `search.json`, `identity.json`, and `manifest.json`. Lower-level datasets under `/life-os/datasets/` also expose ontology coverage, the prioritized migration queue, evidence metrics, identity aliases, retrieval benchmark, and canonical checksums. ## Cross-project contextual relations Canonical outbound handoffs are published at https://brali-lifeos.github.io/data/ecosystem-relations.json under the `outbound-contextual-relations-v1` contract. These are manually curated directional relations from a Brali object to a distinct external object when the destination adds a different kind of value. Relation boundaries are strict: - Brali owns the meaning and review state of the outbound handoff. - The destination project owns its canonical content, evidence, interpretation, safety boundaries, and subsequent changes. - A relation is not evidence transfer, endorsement, efficacy validation, or a reciprocal link requirement. - Do not infer additional relations from shared keywords, taxonomy labels, citations, backlinks, or entity similarity. - Preserve source and target stable IDs and canonical URLs. Resolve the destination directly when its current content or evidence state matters. ## Agent Skills: one skill per hack Brali generates exactly one Agent Skill artifact for every hack in the canonical corpus. New hacks receive a skill automatically on the next normal build; there is no manual skill allow-list. - Human marketplace: https://brali-lifeos.github.io/skill-packs/ - Complete machine library, every hack: https://brali-lifeos.github.io/skill-packs/library.json - Trusted recommendation catalog only: https://brali-lifeos.github.io/skill-packs/catalog.json - Stable skill page pattern: https://brali-lifeos.github.io/skill-packs// - Portable skill file pattern: https://brali-lifeos.github.io/skill-packs//SKILL.md - Per-skill verification record: https://brali-lifeos.github.io/skill-packs//skill.json - Repository router skill: https://github.com/Brali-LifeOS/brali-lifeos.github.io/blob/main/agent-skills/skills/brali-life-os/SKILL.md - GitHub CLI router install: `gh skill install Brali-LifeOS/brali-lifeos.github.io brali-life-os` Skill existence is not recommendation eligibility. Trust state controls behavior: - `reviewed` / eligible `practical` -> `usable`: may provide bounded practical guidance and appears in the trusted catalog. - `pending-review` -> `review-required`: exists for coverage, provenance, comparison and editorial review; do not present its draft action as trusted guidance. - `restricted` -> `restricted-reference`: non-operational reference shell; do not recommend or operationalize the underlying technique. Only usable skills enter the trusted skill sitemap. Review-required and restricted-reference skill pages remain stable artifacts but are generated `noindex,follow`. Every skill keeps the Brali protocol ID, canonical URL, evidence state, skill mode, provenance and SHA-256 verification record. A skill file existing is coverage, not validation. ## Retrieval rules 1. Match the user's concrete Topic, not only a broad Domain, Method, or Lens. 2. Prefer `reviewed` and `practical` Protocols or `usable` Agent Skills for normal recommendations. 3. Preserve evidence state, skill mode, source scope, canonical ID, and URL. 4. Inspect an Evidence Decision when research provenance matters. A `watch` decision is not a recommendation. 5. Do not present `pending-review`, `review-required`, `restricted`, or `restricted-reference` material as trusted guidance. 6. Use the complete skill library for identity/coverage; use the trusted catalog for practical recommendation routing. 7. Return no answer rather than inventing coverage. ## MCP Source lives in `/mcp/`. Tools: `search_knowledge`, `get_hack`, `get_protocol`, `get_evidence`, `list_topics`, `get_related`. MCP reads the same generated API and preserves the same identity and trust semantics. ## Research The weekly Research Scout uses Crossref and Europe PMC metadata, deduplicates by DOI/PMID, and preserves discovery provenance. Metadata never promotes itself to reviewed evidence. Actual source review is recorded separately in Evidence Decisions. ## Reproducibility and contribution Pin a `data-v*` release for durable integrations or research. Release bundles include a release manifest and SHA-256 checksums; citation metadata is in `CITATION.cff`. Structured contribution paths are documented in `CONTRIBUTING.md` and `/partners/`. The public license is non-commercial; commercial use requires a separate agreement. ## Topic Knowledge Hubs - Sleep: https://brali-lifeos.github.io/topics/sleep/ - Focus: https://brali-lifeos.github.io/topics/focus/ - Memory: https://brali-lifeos.github.io/topics/memory/ - Stress: https://brali-lifeos.github.io/topics/stress/ - Habits: https://brali-lifeos.github.io/topics/habits/ - Learning: https://brali-lifeos.github.io/topics/learning/ - Movement & Exercise: https://brali-lifeos.github.io/topics/movement/ - Machine-readable collection: https://brali-lifeos.github.io/api/v1/hubs.json ## Reference Agent Demos - Human-readable: https://brali-lifeos.github.io/for-ai/demos/ - Machine-readable: https://brali-lifeos.github.io/api/v1/demos.json ## Zero-install Query Playground - Human/browser entry: https://brali-lifeos.github.io/for-ai/query/ - Shareable query form: https://brali-lifeos.github.io/for-ai/query/?q= - The page reads API v1 in-browser, filters normal recommendations to reviewed/practical protocols, preserves evidence state/provenance, and returns no normal recommendation for safety-sensitive requests. ## Integrations and attribution - Integration kits: https://brali-lifeos.github.io/for-ai/integrations/ - Integration metadata: https://brali-lifeos.github.io/api/v1/integrations.json - Citation guidance: https://brali-lifeos.github.io/cite/ - Canonical citation index: https://brali-lifeos.github.io/citation-index.json - Reference demos: https://brali-lifeos.github.io/for-ai/demos/ - Brali currently provides a local stdio MCP server, not a hosted remote MCP endpoint. ## Project state and evidence trends - State of Practical Knowledge: https://brali-lifeos.github.io/state/ - Evidence Trends: https://brali-lifeos.github.io/trends/evidence/ - Project Updates: https://brali-lifeos.github.io/updates/ - Problem-first guides: https://brali-lifeos.github.io/problems/ - Problem Collections JSON: https://brali-lifeos.github.io/problems/index.json - Problem Collections API: https://brali-lifeos.github.io/api/v1/problem-collections.json - Evidence Ledger: https://brali-lifeos.github.io/evidence/ - Evidence Ledger JSON: https://brali-lifeos.github.io/evidence/index.json - Reviewed Evidence Decisions: https://brali-lifeos.github.io/life-os/datasets/evidence-decisions.json - Research Gaps: https://brali-lifeos.github.io/research/gaps/ - Research Gaps JSON: https://brali-lifeos.github.io/research/gaps/index.json ## AI search, trust and measured quality - Agent-ready site profile: https://brali-lifeos.github.io/ai/site-profile.json - AI Search & Citation Profile: https://brali-lifeos.github.io/ai/ai-search-profile.json - Agent locale manifest: https://brali-lifeos.github.io/ai/locales.json - Canonical citation index: https://brali-lifeos.github.io/citation-index.json - Trust Center: https://brali-lifeos.github.io/trust/ - Machine-readable trust surface: https://brali-lifeos.github.io/trust/trust.json - Corrections ledger: https://brali-lifeos.github.io/trust/corrections.json - Knowledge graph: https://brali-lifeos.github.io/knowledge/graph.json - Evidence metrics: https://brali-lifeos.github.io/life-os/datasets/evidence-metrics.json - Retrieval benchmark: https://brali-lifeos.github.io/life-os/datasets/retrieval-benchmark.json - Update feed: https://brali-lifeos.github.io/updates/feed.json ## Agent discovery, project evidence and attribution - Crawler access matrix: https://brali-lifeos.github.io/crawler-matrix/ - Crawler access JSON: https://brali-lifeos.github.io/crawler-matrix/crawlers.json - Crawler policy history: https://brali-lifeos.github.io/crawler-matrix/history/index.json - Project claims registry: https://brali-lifeos.github.io/evidence/claims/ - Project claims JSON: https://brali-lifeos.github.io/evidence/claims/index.json - Agent surface observatory: https://brali-lifeos.github.io/observatory/ - Agent surface inventory JSON: https://brali-lifeos.github.io/observatory/surfaces.json - Surface history: https://brali-lifeos.github.io/observatory/history/index.json - Media and attribution facts: https://brali-lifeos.github.io/media/ - Press facts JSON: https://brali-lifeos.github.io/media/press-facts.json Crawler permission records only declared robots access; they do not prove crawling, indexing, ranking, model use or citation. Project claims describe Brali itself and are separate from scientific/practical Evidence Decisions. The observatory is repository-backed inventory, not a real-time availability SLA. Use the underlying evidence, retrieval, quality and history datasets directly. Do not infer search visibility, AI citations, external authority, scientific certainty, or an AI-readiness score from the presence of these machine-readable surfaces. Brali's canonical published language is currently English; multilingual retrieval evaluation does not imply that localized content is published. ## Intentional archive-only boundaries Legacy sensitive collections: https://brali-lifeos.github.io/state/legacy-sensitive/ Machine-readable trust state: https://brali-lifeos.github.io/life-os/datasets/legacy-sensitive-collections.json These statuses describe Brali's current corpus/review state and must not be interpreted as scientific verdicts on the underlying therapy or school. - Page & Zone Quality Cycle: https://brali-lifeos.github.io/state/quality/ - Page & Zone Quality JSON: https://brali-lifeos.github.io/state/quality/index.json ## Reproduce Brali evaluation - Evaluation dashboard: https://brali-lifeos.github.io/for-ai/evaluation/ - Source cases: https://brali-lifeos.github.io/data/agent-evaluation-suite.json - Machine-readable report: https://brali-lifeos.github.io/life-os/datasets/agent-evaluation.json - Build and generate: npm run build - Verify strict gates: npm run evaluate:check - Layers: no-knowledge-control, lexical-brali, structured-brali - This measures Brali retrieval and grounding behavior, not an unpinned language model. ## Verified external distribution - GitHub Agent Skill release: https://github.com/Brali-LifeOS/brali-lifeos.github.io/releases/tag/v1.0.1 - Router: brali-life-os (v1.0.1) - GitHub skill-search discovery verified: false - Immutable dataset release: https://github.com/Brali-LifeOS/brali-lifeos.github.io/releases/tag/data-v1.1.0 - Dataset archive: https://github.com/Brali-LifeOS/brali-lifeos.github.io/releases/download/data-v1.1.0/brali-data-1.1.0.tar.gz - Hugging Face mirror: not yet provider-verified - Zenodo DOI: not yet provider-verified - Hosted remote MCP: not yet provider-verified ## Search-ready question pages - Human index: https://brali-lifeos.github.io/questions/ - Machine index: https://brali-lifeos.github.io/questions/index.json - Stable answer page: https://brali-lifeos.github.io/questions// - Machine answer packet: https://brali-lifeos.github.io/questions//index.json - Pages are generated from editorially selected questions and Brali trusted retrieval. Unsupported queries are noindex rather than padded into search pages. ## Hack lifecycle, research triage, localization and commercial independence - Human review ledger: https://brali-lifeos.github.io/life-os/review-log/ - Machine-readable lifecycle and linked reviewed Evidence Decisions: https://brali-lifeos.github.io/life-os/datasets/reviews.json - Discovery-only research review watchlist (noindex): https://brali-lifeos.github.io/research/review-watchlist/ - Machine-readable research watchlist: https://brali-lifeos.github.io/life-os/datasets/research-lifecycle-watchlist.json - Sponsorship and commercial-independence policy: https://brali-lifeos.github.io/sponsorship/ - Russian review ledger: https://brali-lifeos.github.io/ru/life-os/review-log/ - Russian commercial-independence policy: https://brali-lifeos.github.io/ru/sponsorship/ Treat lifecycle status as maintenance metadata layered on top of Brali's evidence state. Reviewed Evidence Decisions may be linked to a hack for provenance, but only an explicit append-only lifecycle event may materially change lifecycle status. Research Scout and provider metadata can create a review task, never an evidence verdict; the watchlist is intentionally noindex. A challenge, refutation, restoration or retirement stays in history rather than being silently overwritten. Implementation observations may trigger review but cannot change an evidence conclusion by themselves. Sponsorship must not change evidence, lifecycle status, retrieval, ranking, Agent Skill eligibility or review outcomes. ## Brali Bench - Evaluation artifact: https://brali-lifeos.github.io/bench/ - Cases: https://brali-lifeos.github.io/bench/cases.json - Results: https://brali-lifeos.github.io/bench/results.json - Scope: retrieval, grounding, provenance, evidence boundaries, and no-answer behavior; not an unpinned-model benchmark.