Root Agent Skill Framework
A prompt and skill framework for Root Agent, a coordinator Bot on Grok Bot that sizes teams, creates Bots, and delegates work to Leads.
At a Glance
Root Agent description, skills and templates for Grok Bot, released under the MIT License.
Engagement
Available On
Alternatives
Listed Oct 2026
About Root Agent Skill Framework
Root Agent Skill Framework is the MIT-licensed source repository, maintained by MrBekoX, for the description, skills and templates of Root Agent, a coordinator Bot that runs on Grok Bot. It is a prompt and skill engineering project rather than a code project: the texts are copied into Grok Bot or attached to a Bot. The current source version is 0.3.7, with eighteen Root skills and two optional Lead skills.
What It Is
Root Agent acts as the user's single entry point. The user states what they want; Root Agent clarifies the request, chooses a team structure matched to the work, creates the needed Bots, hands management to the right place, verifies delivery, and returns the result with its evidence. It does not do the substantive work itself, except for tier 0 work with no external effects, one deliverable, roughly 10 minutes of effort and no independent-review requirement.
How the tier model works
Root Agent picks a structure from the whole accepted scope. Tier 0 is handled by Root itself, tier 1 uses one Specialist managed directly by Root, tier 2 uses a Lead plus as many Specialists as the work needs, and tier 3 uses one Lead per domain with Root managing across domains. Handoffs to a Lead use a package and require a readback from the verified Lead before work is treated as lead-managed. Root keeps roster creation, independent reviewer selection and scope-expansion decisions.
The skill set
The eighteen skills cover intake, scope control, work queue management, roster management, access probing, team prompting, dispatch, Lead handoff, quality review, blocker escalation, standing workspace records, routine binding, learning, portfolio status, result relay, user reports, deleted-work cleanup and onboarding. Every skill follows the same six sections. Core rules stress evidence before claims, no silent scope narrowing, one owner per stage and one writer per file, and user approval for any action with external effects.
Setup path
Users can add Root Agent from the share link with Add to Grok Bot, self-install from GitHub by attaching the setup prompt to an existing Root Agent conversation, or attach the generated setup instructions file from a source snapshot. Python scripts validate source consistency and regenerate the setup file. The README states that live Grok Bot behavior and the share link configuration have not been verified against this version.
Community Discussions
Be the first to start a conversation about Root Agent Skill Framework
Share your experience with Root Agent Skill Framework, ask questions, or help others learn from your insights.
Pricing
Open Source (MIT)
Root Agent description, skills and templates for Grok Bot, released under the MIT License.
- MIT License
- Eighteen Root skills and two optional Lead skills
- Prompt and skill engineering source: description, skills and templates
- Runs on Grok Bot (usage of Grok Bot itself is separate and not priced in the evidence)
Capabilities
Key Features
- Single entry point coordinator Bot for Grok Bot
- Tier 0-3 structure selection based on accepted scope
- Lead handoff packages with readback gate
- 18 Root skills and 2 optional Lead skills
- Independent review gate with bounded repair loop
- Access probing for read, write and auth separately
- Scope guard with P0, P1 and deferred bands
- Blocker classification and escalation
- Routine binding through the executing owner
- Working records kept in a private workspace
- Measured effort reports and performance cards
- Offline validation scripts and regression tests
