convert-skill-to-teamrun-members
Skill402 starsChanged 7 days agoApache-2.0View on GitHub
Convert an external skill, expert methodology, repository, workflow, or role set into Agent Tower TeamRun members, MemberPresets, rolePrompt documents, Leader or PM Leader orchestration, and optional TeamRun templates. Use when the user asks to turn a skill/repo/method/process into TeamRun roles, team members, presets, prompts, or a reusable Agent Tower team configuration.
In this repository7 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agent-tower-devSkill
- github-desktop-releaseSkill
- publishSkill
- agent-tower-devSkill
- publishSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
38 rules. About 0 are common, 24 are distinctive, and 14 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈24
Distinctive · estimate
This author's own choices. Worth a look.
14
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use this skill to translate a source methodology into an Agent Tower TeamRun team design.
- Distinctive: line 12
Read the source material.
- Distinctive: line 18
Create a TeamRun member only when the capability has a distinct responsibility, clear input/output contract, and meaningful independent …
- Distinctive: line 19
Put routing logic, sequencing rules, and result integration in the Leader or PM Leader prompt instead of creating a member for every …
- Very specific: line 20
Do not memberize reference material, low-value checklists, pure terminology, or work that is better handled by an existing …
- Very specific: line 24
Use the exact capability fields: readRoom, postRoomMessage, mentionMembers, stopMemberWork, markReadyForReview, readFiles, writeFiles, …
- Very specific: line 25
Default expert and specialist roles to triggerPolicy: MENTION_ONLY so the Leader explicitly invokes them.
- Very specific: line 26
Default user-entry Leaders to triggerPolicy: USER_MESSAGES.
- Very specific: line 27
Default specialists to minimum capabilities: usually readRoom and postRoomMessage.
- Very specific: line 28
Use workspacePolicy: none when a role does not need repository access.
- Very specific: line 29
Use workspacePolicy: shared with readFiles for roles that inspect docs/code but do not edit.
- Very specific: line 31
Use sessionPolicy: resume_last for roles that benefit from continuity, and new_per_request for independent audit/check roles.
- Very specific: line 32
Use queueManagementPolicy: own_only for specialists and team_pending only for coordinator roles that must manage team queues.
- Very specific: line 35
Do not duplicate system or Team Room shared protocols in rolePrompt.
- Distinctive: line 38
Keep prompts focused on operating behavior, not full copies of the source methodology.
- Distinctive: line 42
Keep ordinary Leader and domain Leader separate when the workflow is specialized enough to add many domain rules.
- Distinctive: line 43
Put these rules in the Leader prompt: when to call each specialist, required input for each delegation, how to integrate results, when to …
- Distinctive: line 45
Do not let multiple specialists listen to all user messages by default.
- Distinctive: line 48
Prefer Markdown prompt docs first when the team design is still being reviewed.
- Distinctive: line 49
Create or update MemberPresets only after the role prompts and configurations are clear.
- Distinctive: line 50
Create a TeamRun template only when the user explicitly wants a reusable team composition.
- Distinctive: line 51
If configuring through the UI, use browser automation to create presets and then reopen the details to verify saved values.
- Distinctive: line 55
Verify MemberPreset fields in UI or API: names, aliases, provider, policies, capabilities, and rolePrompt source.
- Distinctive: line 56
Verify TeamRun template membership only if a template was requested.
- Very specific: line 57
Check git diff when editing files.
- Distinctive: line 61
When converting product-management skills or methods, prefer this proven pattern unless the source material clearly argues otherwise:
- Very specific: line 63
Use a user-entry PM Leader or Product Owner with triggerPolicy: USER_MESSAGES, workspacePolicy: none, sessionPolicy: resume_last, …
- Very specific: line 64
Use expert PM roles with triggerPolicy: MENTION_ONLY and queueManagementPolicy: own_only.
- Distinctive: line 68
Keep PM experts from replacing engineering roles.
- Distinctive: line 72
Put generic Team Room behavior, visibility rules, result requirements, security rules, and message mechanics in shared protocol, not in …
- Distinctive: line 73
Put domain orchestration, routing, sequencing, result integration, queue handling, and user decision framing in the Leader or PM Leader …
- Very specific: line 74
Put specialist expertise, input expectations, artifact shape, handoff rules, and strict non-goals in each specialist rolePrompt.
- Distinctive: line 75
Put provider IDs, avatars, preset names, aliases, and capability matrices in MemberPreset or TeamTemplate docs rather than burying them …
- Distinctive: line 121
Do not update existing MemberPresets, TeamRun templates, or prompts unless the user explicitly asks.
- Distinctive: line 122
Do not create broad compatibility prompts when a separate domain Leader would be simpler and safer.
- Distinctive: line 123
Do not grant implementation powers to expert advisory roles by default.
- Distinctive: line 124
Do not overwrite project operating files such as AGENTS.md or CLAUDE.md from a source methodology without explicit confirmation.
- Distinctive: line 125
Do not present uncreated members or templates as if they already exist.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Convert Skill to TeamRun Members
Use this skill to translate a source methodology into an Agent Tower TeamRun team design. Keep the output practical: roles, permissions, prompts, orchestration, landing steps, and verification.
Workflow
- Read the source material.
- Inspect the external skill, repo, docs, examples, and any existing prompt files.
- Identify capability modules, expected artifacts, inputs, outputs, handoff points, and explicit boundaries.
- Separate reusable method knowledge from one-off conversation history; do not paste long source text into prompts.
- Decide what should become a member.
- Create a TeamRun member only when the capability has a distinct responsibility, clear input/output contract, and meaningful independent result.
- Put routing logic, sequencing rules, and result integration in the Leader or PM Leader prompt instead of creating a member for every concept.
- Do not memberize reference material, low-value checklists, pure terminology, or work that is better handled by an existing Implementer/Reviewer/Tester role.
- Define each role configuration.
- Specify
name,aliases,providerIdrecommendation if known,workspacePolicy,triggerPolicy,sessionPolicy,queueManagementPolicy, and capabilities. - Use the exact capability fields:
readRoom,postRoomMessage,mentionMembers,stopMemberWork,markReadyForReview,readFiles,writeFiles,runCommands,readDiff,mergeWorkspace. - Default expert and specialist roles to
triggerPolicy: MENTION_ONLYso the Leader explicitly invokes them. - Default user-entry Leaders to
triggerPolicy: USER_MESSAGES. - Default specialists to minimum capabilities: usually
readRoomandpostRoomMessage; addmentionMembersonly when they must hand off work. - Use
workspacePolicy: nonewhen a role does not need repository access. - Use
workspacePolicy: sharedwithreadFilesfor roles that inspect docs/code but do not edit. - Grant
writeFiles,runCommands,readDiff, ormergeWorkspaceonly when the role's normal job requires them. - Use
sessionPolicy: resume_lastfor roles that benefit from continuity, andnew_per_requestfor independent audit/check roles. - Use
queueManagementPolicy: own_onlyfor specialists andteam_pendingonly for coordinator roles that must manage team queues.
- Specify
- Write role prompts.
- Do not duplicate system or Team Room shared protocols in
rolePrompt; assume the system layer injects them. - Include the role's mission, responsibilities, non-responsibilities, inputs, output/result contract, collaboration rules, and verification expectations.
- Make each role aware of its boundaries: it should not impersonate other specialists, implementers, reviewers, testers, or the Leader.
- Keep prompts focused on operating behavior, not full copies of the source methodology.
- Do not duplicate system or Team Room shared protocols in
- Design Leader orchestration.
- Decide whether the existing ordinary Leader is enough or whether a separate domain Leader is cleaner.
- Keep ordinary Leader and domain Leader separate when the workflow is specialized enough to add many domain rules.
- Put these rules in the Leader prompt: when to call each specialist, required input for each delegation, how to integrate results, when to ask the user, and when to hand off to engineering, review, testing, or audit.
- The Leader should coordinate and synthesize; it should not personally produce complete expert artifacts that belong to specialist roles.
- Do not let multiple specialists listen to all user messages by default.
- Land the artifacts.
- Prefer Markdown prompt docs first when the team design is still being reviewed.
- Create or update MemberPresets only after the role prompts and configurations are clear.
- Create a TeamRun template only when the user explicitly wants a reusable team composition.
- If configuring through the UI, use browser automation to create presets and then reopen the details to verify saved values.
- Verify.
- Review generated prompts for duplicated system protocols, vague boundaries, overbroad permissions, and missing result contracts.
- Verify MemberPreset fields in UI or API: names, aliases, provider, policies, capabilities, and rolePrompt source.
- Verify TeamRun template membership only if a template was requested.
- Check git diff when editing files; it should contain only the intended prompt/skill/template artifacts.
Product-Team Defaults
When converting product-management skills or methods, prefer this proven pattern unless the source material clearly argues otherwise:
- Use a user-entry
PM LeaderorProduct OwnerwithtriggerPolicy: USER_MESSAGES,workspacePolicy: none,sessionPolicy: resume_last,queueManagementPolicy: team_pending, and coordinator capabilities such asreadRoom,postRoomMessage,mentionMembers, and optionallystopMemberWork. - Use expert PM roles with
triggerPolicy: MENTION_ONLYandqueueManagementPolicy: own_only. - Product strategy and discovery roles usually need no repository access.
- Execution/specification roles may use
workspacePolicy: sharedplusreadFileswhen they need project docs or code context. - Shipping/audit roles may use
workspacePolicy: shared,readFiles, andreadDiff; they should not write, run commands, or merge by default. - Keep PM experts from replacing engineering roles. They prepare strategy, discovery, PRD, acceptance criteria, audit findings, and handoff inputs; Implementers, Reviewers, and Testers do their own work.
Prompt Placement
- Put generic Team Room behavior, visibility rules, result requirements, security rules, and message mechanics in shared protocol, not in individual role prompts.
- Put domain orchestration, routing, sequencing, result integration, queue handling, and user decision framing in the Leader or PM Leader prompt.
- Put specialist expertise, input expectations, artifact shape, handoff rules, and strict non-goals in each specialist
rolePrompt. - Put provider IDs, avatars, preset names, aliases, and capability matrices in MemberPreset or TeamTemplate docs rather than burying them only inside role prompts.
Output Template
## Role Inventory
- Role name: purpose, main artifact, why it should or should not be a TeamRun member.
## Recommended MemberPreset Configuration
| Role | triggerPolicy | workspacePolicy | sessionPolicy | queueManagementPolicy | Key capabilities |
| --- | --- | --- | --- | --- | --- |
| ... | ... | ... | ... | ... | ... |
## Leader Orchestration
- User-entry member:
- When to call each specialist:
- Required delegation input:
- How to integrate results:
- Engineering/review/testing/audit handoff:
## Prompt Artifacts
- Files to create or update:
- Role prompt boundaries:
- Shared protocol content that should not be duplicated:
## Landing Steps
- Markdown prompts:
- MemberPresets:
- Optional TeamRun template:
- UI/API verification:
## Risks and Open Questions
- Missing source context:
- Permission risks:
- Roles that may overlap:
- User decisions required:Guardrails
- Do not update existing MemberPresets, TeamRun templates, or prompts unless the user explicitly asks.
- Do not create broad compatibility prompts when a separate domain Leader would be simpler and safer.
- Do not grant implementation powers to expert advisory roles by default.
- Do not overwrite project operating files such as
AGENTS.mdorCLAUDE.mdfrom a source methodology without explicit confirmation; propose a patch or separate prompt artifact instead. - Do not present uncreated members or templates as if they already exist; distinguish design, docs, UI/API configuration, and verified runtime state.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

