Skills

Usable AI operating skills.

A public library of SKILL.md files for moving work from signal to evidence, review, handoff, and controlled action. Each skill includes inputs, workflow steps, output contracts, and guardrails.

Open the skills repo Featured skills Full library

Operating Model

Signal to source to evidence to review to action.

The skills are meant to be chained, but each file can also stand on its own as a reusable workflow contract.

Signal intake Source review Evidence contract Review queue Handoff

Featured Skills

The seven most useful public skill files.

These are the clearest examples of the operating pattern: collect context, preserve evidence, route review, and prepare the next move.

SKILL.md Signal intake

Collects source changes, market movement, account activity, and operating signals into a reviewable queue.

Creates the first filter: what changed, why it may matter, and whether it deserves action. Open skill file SKILL.md Source review

Turns source documents, links, and notes into grounded briefs with clear evidence boundaries.

Separates facts, interpretations, unsupported claims, open questions, and the source material behind them. Open skill file SKILL.md Evidence contract

Carries decision, evidence, confidence, risks, and source URLs forward between workflow stages.

Prevents later AI steps from re-inventing facts from a thin summary or losing provenance. Open skill file SKILL.md Review queue routing

Moves records through explicit states such as needs review, rejected, confirmed, draft ready, and approved.

Keeps ambiguous or risky transitions human-owned while repeatable work moves forward. Open skill file SKILL.md CRM handoff

Converts unstructured client communication into structured CRM updates, follow-ups, and account memory.

Designed around field ownership, next steps, source context, and review before write. Open skill file SKILL.md Founder brief

Condenses company chat, updates, and operating context into founder-ready briefs.

Pulls decisions, risks, priorities, and unresolved questions out of high-volume collaboration surfaces. Open skill file SKILL.md Decision review

Compares evidence, constraints, owners, and timing before recommending the next operating move.

Keeps the output practical: what matters, what is uncertain, who owns it, and what happens next. Open skill file

Full Library

Twenty-five skill packs grouped by workflow stage.

The full set covers intake, source review, prioritization, drafting, handoff, sync boundaries, logging, reporting, and safe control surfaces.

Intake And Evidence

Skills for turning messy source material and observed changes into structured, reviewable records.

Source registry Defines approved sources, source types, query context, and parser expectations. Source scraping Collects public evidence from approved sources while preserving provenance. Signal intake Turns observed source changes into a reviewable queue with confidence and next action. Context intake Normalizes messy names, URLs, notes, snippets, and metadata into structured records. Source review Separates facts, interpretations, unsupported claims, open questions, and source evidence. Evidence contract Carries decisions, evidence, confidence, risks, and source URLs between workflow stages.

Review And Prioritization

Skills for deciding what is real, what matters, and what is allowed to move forward.

Dedupe and history Prevents repeated records while preserving useful source and scan history. Identity review assist Checks whether an entity is real and correctly matched before enrichment. Review queue routing Moves records through explicit states before enrichment, drafting, handoff, or action. Enrichment Adds context, fit signals, evidence-backed claims, and risks to confirmed records. Scoring and prioritization Turns evidence into priority, fit, route, and review decisions. Decision review Compares evidence, constraints, owners, timing, and downstream impact before action.

Drafting And Handoff

Skills for turning reviewed context into drafts, briefs, CRM handoffs, and operator-ready outputs.

Draft preparation Produces evidence-grounded drafts without sending anything. Draft quality review Checks drafts for supported claims, weak reasoning, and approval readiness. CRM handoff Converts communication into CRM-ready notes, field recommendations, tasks, and routing. CRM handoff validation Validates approval gates and required fields before downstream CRM sync. Founder brief Turns chat, notes, and operating context into founder-ready decisions and next actions. Reporting and digesting Turns raw run output into operator-ready summaries, exceptions, and decisions needed.

Execution Boundaries

Skills for keeping automation auditable, reviewable, and safe before external or state-changing actions.

MCP tool-call wrapping Wraps approved payloads for connector-style execution without bypassing gates. Sync reconciliation Compares review-state records against downstream state before live writes. Structured logging Records workflow actions and decisions in auditable machine-readable form. Context filing Files notes, lessons, and decisions into the smallest correct source of truth. Secret-safe handoff Produces setup or review packets without exposing credentials. No-send boundary Keeps drafting separate from outbound transmission and approval. Control surface Defines safer interfaces for previewing, reviewing, and triggering automation.