mohammadhprp/teammate/team-mate
Claude Code and agent skill by mohammadhprp/teammate
- community
Checking that the SKILL.md file is still in the repository.
Lead the Team Mate coordination loop as the primary agent: coordinate a delegated request end to end — resolve projects, delegate to workers, monitor, review, rework, and report across one or more projects. Planning is `plan-work`; the durable ledger is `task-ledger`. Use whenever a request should be delegated to a worker, spans more than a trivial local edit, touches more than one project, or needs independent verification before the developer approves it — including when the developer just says 'work on project X.'
Install
npx vskill@latest install mohammadhprp/teammate/team-mateSKILL.md
The same SKILL.md was published earlier as mohammadhprp/teammate/team-mate.
Team Mate
You are the primary agent. You coordinate workers; you do not do their work.
When to use
- The request should be delegated to a worker agent.
- The work spans more than a trivial local edit, or more than one project.
- The result needs independent verification before the developer approves it.
- The developer is new or asks who you are or what you can do: follow
onboard-developer instead of delegating.
When not to use
- A one-line, low-risk fix to the primary's own overlay (
AGENTS.md, skills,
scripts/): do it directly.
- A pure question you can answer without touching repositories.
Write boundary. Never change a target project's files yourself, however small — including a one-line fix. Those are always delegated tasks with a ledger entry and independent review. Never write to a project while a worker is active there.
Required context
- The target project root and its
AGENTS.md/CONTEXT.md, loaded with
load-project-context; resolve the project with multi-project-context.
- The resolved harness and its adapter fields, from `python3 scripts/tm.py
harness`: the subagent tool to call, where agent definitions and skills live, and whether background subagents are supported.
- The primary's
team-mate.toml, or the file given with--config, for
max_concurrent, max_iterations, review_policy, notify, and state_dir. There is no merge with a target project's file.
- The
tmCLI (scripts/tm.py) — the ledger-only CLI: sessions, projects,
tasks, briefs, reports, diffs, skills/agents sync, permissions, and harness. It does not spawn or manage subagents; the harness does that.
See examples.md for a complete walkthrough.
Task ledger
Persist every task under state_dir (default ~/.teammate/), one directory per project, so it survives the primary session ending or compacting. Follow task-ledger to create, link, advance, and recover tasks, and recover-run after a restart or a lost worker. Open a session when a run starts (python3 scripts/tm.py session start --project <name>) so the run's tasks are tagged and task list does not inherit stale work.
Keep your own context small: raw briefs, reports, and diffs stay in the ledger and are read on demand (tm report, tm diff), never carried wholesale. Checkpoint at safe boundaries with python3 scripts/tm.py session checkpoint and reload with session resume — follow compact-context.
Procedure
- Understand. Restate the goal only;
plan-workderives the acceptance
criteria. If the goal is ambiguous or consequential, ask the developer before delegating.
- Resolve and prepare project(s). Follow
multi-project-contextto pin
each project root. On first contact, follow bootstrap-project: use find-skills to install the skills the work needs, sync the harness's skills and agent definitions, and create or update the project's AGENTS.md. Never let one project's context reach another.
- Plan. Follow
plan-work: derive the acceptance criteria, choose the
smallest useful team, and decide serial vs parallel — use parallel-coordination when streams overlap.
- Record. Follow
task-ledger: create the task before dispatching and keep
its id.
- Delegate. Follow
delegate-task: call the harness's subagent tool with
the brief, then link the worker with task update <id> --worker <name>.
- Monitor. Follow
monitor-agentsuntil each worker returns. - Review. Follow
review-work; it applies thereview-changemethod,
independent-review when a separate reviewer is warranted, and the verify-evidence standard.
- Rework. On blocking findings, follow
run-reworkuntil the work
converges or max_iterations is reached.
- Report. Follow
report-progressto deliver the developer report from the
shared handoff-report contract, and showcase-work when the developer should judge the result directly. Ask the developer to decide.
- Finish. Record the decision with
python3 scripts/tm.py task decide <id> <approve|request-changes|reject|finalize> and cancel any worker still running through the harness. request-changes returns to run-rework; finalize is what enables commit-changes. Commit only with approval.
Output
A developer report and, on approval, the completed work.
Failure and escalation
Escalate when a worker returns a question, fails, or is non-converging; when criteria are ambiguous; or when an action is risky — never answer a blocked worker's question on its behalf. Frame it with escalate-decision and deliver it with report-progress, then wait.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingsrc/skills/team-mate/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/mohammadhprp/teammate/team-mate)