agentlas-ai/agentlas-os-team-builder-packaging
Use when generating or auditing a multi-role agent team package with orchestrator, PM Soul, Memory Curator, Policy Gate, workers, eval, QA, handoffs, and runtime adapters.
npx skills add https://github.com/agentlas-ai/Agentlas-OS --skill team-builder-packaging
docs/builder-interview-research-gate.md before writing the roster:ask an 8-12 question first batch, research official sources, similar agent
repositories or comparables, academic/professional theory, and plugin docs,
compare tool/plugin choices, and write the domain-expert synthesis plus
prompt-performance contract.
docs/builder-interview.md, docs/research-sources.md,docs/tool-selection.md, docs/domain-expert-synthesis.md,
docs/prompt-performance-contract.md, and
.agentlas/capability-eval-plan.json unless explicitly creating a minimal
private scaffold.
.agentlas/global-commands.jsonand runtime command files. Do not expose worker commands unless requested.
9b. Declare the execution graph in manifest.json. This is what the Hub runtime
reads to build the team; a package that omits it is published, charged for as
a team, and can never be called:
{
"entrypoints": { "orchestrator": "agents/00-orchestrator/agent.md" },
"roster": ["agents/10-<role>/agent.md", "agents/20-<role>/agent.md"]
}
Write both keys explicitly. The runtime also accepts older spellings
(entrypoint, orchestrator, entry; workers, members, team) and can
derive the roster from agents/<name>/agent.md, but relying on that leaves
the team's shape implicit and it drifts. Note that entry in agentlas.json
is the PACKAGE entrypoint, not the team manager — never reuse it for that.
10. Emit runtime adapters and package verification.
11. Run scripts/verify-team-package.sh <package-root> before reporting
completed. If it fails, do not hand off a result; correct the package by
adding an orchestrator/HQ plus company-blueprint topology or by collapsing
it to a valid single-agent package.
Return team_topology, nodes, edges, memory_architecture, gates,
runtime_adapters, global_commands, and verification.
Take agentlas-ai/agentlas-os-team-builder-packaging from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.