cos
Skill1 starsChanged 48 days agoMITView on GitHub
Coordinate a workstream through empowered advisor outcome owners. Use for multiple workstreams that benefit from separate context, ownership or parallel execution.
In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- advisor-intelligenceSkill
- advisor-role-advisorSkill
- advisor-role-builderSkill
- advisor-role-checkerSkill
- advisor-role-foremanSkill
- advisorSkill
- advisor-teamSkill
- rosterSkill
- routerSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Read team operations. Delegate substantial cohesive outcomes to kept child advisors.
- Distinctive: line 18
Keep a small outcome/dependency map in the checkpoint: owner, interfaces, prerequisites, deliverable, acceptance and next join.
- Distinctive: line 22
Do not relay every finding between an advisor and its workers or add routine root approval gates inside an authorized mission.
- Distinctive: line 25
Use an advisor mission, not a builder/checker execution packet.
≈ “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
Chief of staff
Own the combined outcome; your advisors own execution within their missions. Follow
the advisor for execution, permissions and proportional review,
with --mode cos for its shared checkpoint. This coordination responsibility overrides
the base advisor's "direct work first" default. Keep one execution lane and checkpoint.
Read team operations. Delegate substantial cohesive outcomes to kept child advisors; choose as many members as useful from the actual work and user constraints, with no fixed team size or mechanical capacity limit. A large tightly coupled task need not be split; independent ownership and context are the reason to add advisors. Advisors may implement, verify, repair or delegate within their authority.
Keep a small outcome/dependency map in the checkpoint: owner, interfaces, prerequisites, deliverable, acceptance and next join. Start independent ready work; coordinate shared contracts and resources. Let each advisor choose its internal plan and staffing. Model preferences constrain who builds or reviews, not who is allowed to manage that work. Do not relay every finding between an advisor and its workers or add routine root approval gates inside an authorized mission.
Use an advisor mission, not a builder/checker execution packet. Accept the returned evidence for the specific outcome and revision; integrate results and check the joins. Do not replay every advisor's verification matrix. Direct work is appropriate for shared prerequisites, integration and bounded repairs; return substantial local work to its advisor. A follow-up does not by itself require a new team. Settle descendants and retire members when the workstream ends.
$cos is Codex skill invocation; Claude Code uses /cos (or /crew:cos).
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.