83 canonical skills. Every one documented.

Enforces Pre-Code Gate, model roles, preflight resolution, mid-session handoff, execution logging, and feedback capture. Reads .project-ai/ authority pack when available.

View full spec →

Applies enterprise baseline — all skills, rules, MCP config, testing standards to any machine or project.

View full spec →

Enforces centralized adapter-layer access for third-party SDKs, including singleton factories, typed helper contracts, and no-direct-import enforcement guard rails.

View full spec →

Defines standardized I/O contracts for inter-skill communication. Establishes shared context bus (.project-ai/skill-outputs/), output format, input declarations, and handoff protocol.

View full spec →

Manages reactive skill chaining. Monitors skill outputs for signals and recommends or invokes follow-up skills. Includes built-in triggers for security escalation, quality chains, audit chains, and operations.

View full spec →

Resolves skill execution dependencies at runtime. Reads the canonical registry, builds dependency graphs, detects cycles, and produces ordered execution plans with prerequisite and recommendation chains.

View full spec →

Designs policy gates and approval workflows for AI-assisted development. Defines which skills, workflows, tools, or actions require approval, escalation, or blocking based on environment, risk, and role.

View full spec →

Defines role-based access to skills, workflows, tool categories, and privileged operations. Builds allowlists, denylists, environment scoping, and approval mappings for teams.

View full spec →

Session-scoped cost estimation, budget tracking, and spend forecasting for AI agent usage. Estimates token consumption from observable work, projects burn rate against a budget, and reports at session end.

View full spec →

Estimates total project cost-to-completion across IDE plan tiers. Analyzes project scope, remaining work, and target timeline to recommend the optimal plan and predict upgrade/downgrade triggers.

View full spec →

Mandatory output verification for agent-generated reports, audits, and statistical summaries. Extracts numerical claims, re-executes underlying queries, verifies aggregation integrity, checks methodology consistency, enforces attribution confidence levels, and corrects discrepancies before delivery.

View full spec →