# Telos Brain, AI Skills Directory These skills help you and your AI agent to work with Telos. You can drop this URL into any LLM and it will be able to find and use our skills — helping you to engage with us better. ## BRA101: Telos Brain Core Concepts https://www.telosbrain.com/skills/BRA101?v=2 The building blocks of a Telos Brain, how they relate, and the two authoring styles used to define them — the essential mental model before authoring or deploying a brain. ## BRA102: Company Brain Model https://www.telosbrain.com/skills/BRA102?v=2 The foundational model of what a company brain is — its DNA genome, the brain-vs-database distinction, and the three learning loops that keep it alive and compounding. ## BRA103: What Is a Skill Book https://www.telosbrain.com/skills/BRA103?v=1 What a Skill Book is — a self-contained folder of categorised skills managed by skillbook.yml, portable between brains, with codes that let skills reference each other for progressive disclosure. Use when explaining skill books, placing a skill in a book, or deciding what belongs in skill content. ## BRA104: Getting Started https://www.telosbrain.com/skills/BRA104?v=3 Interactive onboarding interview to configure a brain schema from scratch — entity type, unit of work type, blueprint categories, and skill categories. Use after brain init, before the first deploy. If the host application already exists, use BRA211 (auto-build) instead. For running the local Docker stack, use BRA106. ## BRA105: Brain Principles https://www.telosbrain.com/skills/BRA105?v=3 High-level principles for organising a Telos Brain — skills, tools, configuration, memory, and LLM budgets. Always inject when triaging learnings or editing the brain. ## BRA106: Local Development https://www.telosbrain.com/skills/BRA106?v=1 How to run Telos Brain locally with the CLI and Docker — start, stop, deploy, credentials, and pointing connectors at a host app via host.docker.internal. Use when developing a brain or a host application against a local Brain stack. ## BRA201: Authoring a Telos Brain Schema https://www.telosbrain.com/skills/BRA201?v=10 How to setup a brain schema using yml and markdown ## BRA202: Environment Variables, Secrets & API Keys https://www.telosbrain.com/skills/BRA202?v=5 How a brain's .env variables are uploaded, encrypted and stored; the well-known "system" keys the platform recognises (LLM provider keys, the brain API key); how to inject a stored secret into an api tool's outbound request — as an HTTP header, a query parameter or a JSON body field; how connector credential values relate to the same store; and how connector url-env and parameter secret: bindings resolve named variables from this store. ## BRA203: Schema System Tools https://www.telosbrain.com/skills/BRA203?v=2 The in-brain system tools that let a running brain inspect, edit, and create its own configuration-as-code schema — list_schema_files, search_schema_files, get_schema_file, update_schema_file, create_skill and create_schema_file. Covers what each does, how to wire them into a workflow, the schema file model they expose, and the safety guarantees (brain scoping, exact-one-match edits, automatic versioning, create paths). ## BRA204: Template Tag Taxonomy https://www.telosbrain.com/skills/BRA204?v=2 Canonical reference for double-curly-bracket template tags used in workflow Instructions and tool response-markdown / error-markdown. Covers the input scope from Execution API variables, workflow-tool / run_workflow parameters, and input-tools mappings. ## BRA205: Cloning a Brain https://www.telosbrain.com/skills/BRA205?v=2 How to deep-clone a brain instance via the Management API — POST /brains/{instance}/clone — including what is copied, what is excluded, environment-variable overrides, and the one-time API key response. ## BRA206: Updating a Brain from a Template https://www.telosbrain.com/skills/BRA206?v=2 How to update an existing brain's configuration from a source (template) brain via the Management API — POST /brains/{targetInstance}/update-from/{sourceInstance} — including version precedence, what is synced, embeddings, and the DeploySummary response. ## BRA207: Learning Eval Workflows https://www.telosbrain.com/skills/BRA207?v=2 How to author TRIGGERED learning-eval workflows that grade a completed unit of work or workflow run, inject telemetry via template tags, persist a 0–100 score with set_run_grading, and create inbox learnings with create_inbox_entry — including manual Run eval and automatic trigger modes. ## BRA208: Designing Skill Books https://www.telosbrain.com/skills/BRA208?v=1 How to design a Skill Book's category structure and skill number ranges so the AI can place knowledge correctly. Use when creating a new skillbook, proposing or reviewing categories, writing category descriptions, choosing category indexes (100, 200, 300…), or assigning skill codes within a range. ## BRA209: Connectors https://www.telosbrain.com/skills/BRA209?v=2 How to author connector YAML files for external services (OAuth 2, API key, none, or caller-jwt). Covers file layout, brain-compose registration, optional platform type (e.g. elevenlabs), parameter declarations vs secret storage, url vs url-env, parameter secret: bindings, deploy behaviour, and worked examples. ## BRA210: LLM Providers and Models https://www.telosbrain.com/skills/BRA210?v=2 Supported AI providers for workflow runs, the provider/model string format, example model codes, credential variable names, and which ConversantSettings apply per provider. ## BRA211: Auto-Building a Brain https://www.telosbrain.com/skills/BRA211?v=1 End-to-end process for auto-building a brain schema from an existing application — entity and unit of work, skill-book categories, workflows, tools, Execute API plumbing, and bidirectional secrets. Use instead of the BRA104 interview when the host app already exists. Load BRA105 before starting. ## BRA212: Managing LLM Costs https://www.telosbrain.com/skills/BRA212?v=1 How to keep LLM spend down in a Telos Brain — aim for 80% cache reads (or turn on automatic caching), convert JSON tool data to markdown or CSV, treat tool definitions as mini-skills to cut retries, compact older context, pick a cheaper model (Grok), enforce token budgets plus daily spend and run/turn caps, and use progressive disclosure so the agent discovers skills and tools as needed. Use when authoring or reviewing workflows, tools, or brain-compose.yml for cost, tokens, cache, compaction, or model choice. ## BRA401: Execution API: Authentication & Conventions https://www.telosbrain.com/skills/BRA401?v=6 How to authenticate with the Telos Brain Execution API, how tenancy is resolved, the brain identity endpoint, and the common conventions (status codes, error format, content types) that apply across all endpoints. ## BRA402: Execution API: Entities & Units of Work https://www.telosbrain.com/skills/BRA402?v=5 How to create and manage entities and units of work via the Execution API — including appending context and data logs, and completing a unit of work to trigger the learning eval. ## BRA403: Execution API: Workflow Execution & Telemetry https://www.telosbrain.com/skills/BRA403?v=4 How to list a brain's workflows (with pending inbox-task counts), run them synchronously (SSE streaming) or asynchronously (fire-and-forget with callback), pass optional run variables for {{input.*}} template tags and pre-called input-tools (BRA409), hold a multi-turn chat session against a single run, stop an in-flight turn without closing the session (POST /runs/{id}/stop), and retrieve run telemetry and unit-of-work telemetry via the Execution API — including model, thinking mode and turn totals on run telemetry. Documents when a run settles Failed (max-turns exhausted, final max_tokens with no retries). Also notes when a Completed, Failed, or AwaitingInput run becomes eligible for learning evals (BRA207). Documents SSRF rules for async callbackUrl (allowed-callback-domains). ## BRA404: Execution API: Inbox Entries & Tasks https://www.telosbrain.com/skills/BRA404?v=2 How to create, list, read and update inbox entries and their tasks via the Execution API — the learning-signal intake surface. Covers the entry and task lifecycles, inbox trigger matching (entry create vs task auto-run), learning mode qualifiers, listing by status, adding entries, and updating an entry's or task's status and content (including approving a task to run its linked workflow). In-brain evals should prefer the create_inbox_entry system tool (BRA405) over HTTP. ## BRA405: Inbox System Tools https://www.telosbrain.com/skills/BRA405?v=2 The in-brain system tools for operating the learning-signal inbox — create_inbox_entry, list_inbox_entries, get_inbox_entry, update_inbox_entry, list_inbox_tasks, add_inbox_task and update_inbox_task. All identity is by 8-character reference (never UUID); workflows are selected by code. Covers how PENDING tasks auto-run from the linked workflow's inbox trigger (see BRA404). For persisting a run quality score, see BRA406 (set_run_grading). For listing Organisation members and assigning tasks by email, see BRA408 (list_users, assign_task_to_user). ## BRA406: Run Grading System Tool https://www.telosbrain.com/skills/BRA406?v=2 The set_run_grading system tool — how eval workflows persist a 0–100 quality score on a WorkflowRun (and optionally link an InboxEntry) so the admin console can show traffic-light grades and grading trends. Identity is by 8-character run reference from {{run.reference}}. ## BRA407: Unit of Work Context and Data System Tools https://www.telosbrain.com/skills/BRA407?v=2 The in-brain system tools for recording and reading Unit of Work telemetry — add_uow_context, add_uow_data, update_uow_data, list_uow_context, list_uow_data, get_uow_context and get_uow_data. All identity is by 8-character reference (never UUID). Tools operate only on the harness-injected active Unit of Work. For the Execution API HTTP surface, see BRA402. ## BRA408: Organisation User System Tools https://www.telosbrain.com/skills/BRA408?v=2 The in-brain system tools for discovering Organisation members and assigning InboxTasks — list_users and assign_task_to_user. Scoped to the Brain's Organisation via harness-injected BrainId. list_users returns CSV of name, email, and free-text role profile; assign_task_to_user sets InboxTasks.AssignedTo by member email. For creating a task already assigned, see add_inbox_task assigned_to in BRA405. ## BRA409: Workflow run variables & pre-called input-tools https://www.telosbrain.com/skills/BRA409?v=2 How harness apps pass string variables into a workflow run via the Execution API, how those values surface as {{input.*}} template tags, and how workflows declare input-tools to call tools automatically at startup using those variables. Companion to BRA403 (run endpoints), BRA204 (input scope), and BRA201 §8.0a (YAML schema). ## BRA410: Execution API: File Transcription https://www.telosbrain.com/skills/BRA410?v=2 How to extract text from uploaded files via the Execution API POST /transcription endpoint — supported types, request shape, response envelope, error conditions, and how image transcription resolves its vision model from Brain settings and environment secrets. ## BRA411: Skill Discovery System Tools https://www.telosbrain.com/skills/BRA411?v=2 The in-brain system tools for discovering and loading skills — list_skills, find_available_skills and get_skill. list_skills returns the full SkillBook catalogue as CSV (code, name, description); find_available_skills ranks by semantic relevance; get_skill loads full instructions (and may promote tools).