New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Team

by shenysun in shenysun/skills-manager

Run the four-role Herdr team (pm, po, dev, qa panes) that ships one feature of this repo.

MITAuto-check passedAgent Workflows

Install Team

skills CLI
$ npx skills add shenysun/skills-manager --skill team -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install shenysun/skills-manager team --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/shenysun/skills-manager.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/team .claude/skills/team && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
team
GitHub stars
197
Token cost
~2.7k tokens
SKILL.md length
1,557 words
Files
2 (incl. references)
Skills in repo
2
Repo updated
First seen
Licence
MIT

At a glance

Run the four-role Herdr team (pm, po, dev, qa panes) that ships one feature of this repo.

  • Works in 3 steps: Verify test "${HERDR_ENV:-}" = 1;… → Confirm the cwd is this repo's root —… → Never prompt an agent that is not idle…
  • Agent Workflows work in your project
  • SKILL.md covers Orchestrator mode, Role mode, Scenarios and Findings
  • Calls pnpm and git

What it does

Team is an agent skill from shenysun/skills-manager. Run the four-role Herdr team (pm, po, dev, qa panes) that ships one feature of this repo. In the main session the user runs /team as the orchestrator (up, status, go, down); in a role pane the orchestrator prompts /team role: pm|po|dev|qa. Requires HERDRENV=1.

Its SKILL.md is about 2.7k tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files, including reference files (for example `references/orchestration.md`).

It sits in Agent Workflows. The licence is MIT.

When your agent uses it

  • Agent Workflows work in your project

Example prompts

  • “/team”

Workflow steps

3 steps, taken from the first numbered list in SKILL.md.

  1. Verify test "${HERDR_ENV:-}" = 1; otherwise say you are not inside Herdr and stop.
  2. Confirm the cwd is this repo's root — role panes must start there to load .claude/skills/.
  3. Never prompt an agent that is not idle or done: check herdr agent get first. A blocked role is waiting at a question or approval UI…

What it can do on your machine

Read from SKILL.md and the folder at commit 3cf755f. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • pnpm
    • git

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use pnpm and git, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Team loads about 2.7k tokens when it runs, and up to ~4k if it reads all its reference files. Until then it costs about 67 tokens; SKILL.md has 1,557 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~67
When it runs · the whole SKILL.md, loaded when a task matches
~2.7k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~4k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from shenysun/skills-manager at commit 3cf755f, republished under its MIT licence (© shenysun). 1,557 words, ~2,739 tokens.

Download SKILL.mdSave it as .claude/skills/team/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
team
description
Run the four-role Herdr team (pm, po, dev, qa panes) that ships one feature of this repo. In the main session the user runs /team as the orchestrator (up, status, go, down); in a role pane the orchestrator prompts /team role: pm|po|dev|qa. Requires HERDR_ENV=1.
disable-model-invocation
true

Team

Four roles in Herdr panes ship one feature: pm grills and publishes the spec, po answers the grilling and rules on defects on the user's behalf, dev implements one ticket per invocation, qa derives and executes the acceptance checklist. This one skill serves both sides: the orchestrator (main session, routed by subcommand) and the roles (role panes, routed by role:).

Roles never share context. Files are the interface: every artifact lives in .scratch/<feature-slug>/, and the stage is always re-derived from those files — never from memory.

Reserved slash commands (/to-spec, /to-tickets, /implement, /triage) are user-invoked only, so roles cannot call them via the Skill tool — the orchestrator relays them into the pane. Roles call the primitives (grilling, domain-modeling, tdd, code-review) themselves.

Orchestrator mode

You are not a fifth worker: you create the layout, route turns, and surface escalations to the user. For exact herdr command sequences, prompt texts, and the team.md format, load references/orchestration.md.

Guards
  1. Verify test "${HERDR_ENV:-}" = 1; otherwise say you are not inside Herdr and stop.
  2. Confirm the cwd is this repo's root — role panes must start there to load .claude/skills/.
  3. Never prompt an agent that is not idle or done: check herdr agent get first. A blocked role is waiting at a question or approval UI — never re-prompt it.
up <feature-slug> [one-line idea]
  1. If .scratch/<slug>/team.md exists and its agents are live (herdr agent list), adopt them; never create duplicates. Abort if any team name is live but not ours.
  2. Otherwise create the layout and agents per orchestration.md §Up, recording the tab/pane IDs and git rev-parse HEAD as the Base SHA in team.md.
  3. Send the PM kickoff prompt with --wait --timeout 120000.
  4. Tell the user: PM will interview via PO, then run /team go. Mention the cost: this runs four parallel Claude sessions.
status
  1. herdr agent get each of pm/po/dev/qa.
  2. Derive the stage from files (§Stage). Grep Status: lines; do not read whole files.
  3. Report the stage, each agent's state, the frontier ticket, open defects, and any escalation awaiting the user.
go
  1. Derive the stage (§Stage) and perform only that stage's action. Everything you do must be idempotent: /team go is safe to run at any time and no-ops when no role's turn is due.
  2. Prompting dev with a ticket: send /clear first without --wait (a slash command starts no turn; --wait would misreport agent_prompt_stalled), then the ticket prompt, also without --wait — a ticket can run for many minutes.
  3. Report what you did, and what the user is expected to do next.
down

Ask for confirmation. Send /exit to each agent (no --wait), close only the tab recorded in team.md, then list the leftover .scratch/<slug>/ paths. Never close panes or tabs you did not create.

Stage

Derive from .scratch/<slug>/, first match wins. The same table serves status and go.

ConditionStageAction
no team.mdnot-startedsuggest /team up
no spec.mdpm-interviewrun the Answering loop (orchestration.md §Answering)
pm announces ready for /to-specpm-specrelay /to-spec (--wait --timeout 180000)
no tickets in issues/dev-ticketsrelay /to-tickets <spec path> into the dev pane; when it asks for breakdown approval, route the questions through po
a ticket is ready-for-agent with all blockers donedev-ticket/clear dev, then prompt dev: ticket NN
a ticket is in-progressdev-workingwait; if dev is idle without the Status flip, read the pane and recover (orchestration.md §Recovery)
all tickets done, no acceptance.mdqa-kickoffprompt qa: kickoff
acceptance.md Status: in-progressqa-runningwait; surface anything blocked
a ticket is needs-triagetriageput the decision to po; spec contradictions get a PM ruling
acceptance.md Status: passed, no open ticketsdonesummarize (commits, scenarios, defects); suggest /team down
anything elsestuckreport the file state vs agent states to the user and stop
Blocked roles

A blocked role (or an agent_blocked reply) is waiting at a question or approval UI — never re-prompt it. Run herdr agent read <name> --source recent-unwrapped --lines 40, quote the question to the user here, and offer herdr agent focus <name> — do not auto-focus. Then stop: the user answers in the pane and re-runs /team go.

Role mode

Triggered by /team role: <pm|po|dev|qa> feature: <slug> plus role-specific arguments. Read only your own section; the other roles' files are read-only to you.

PM

Your deliverable is exactly one file: .scratch/<feature-slug>/spec.md. You never write tickets or code.

  1. Load context: read CONTEXT.md for vocabulary and the ADRs touching the area.
  2. Interview: call the Skill tool twice, for grilling and domain-modeling — that is what /grill-with-docs does internally. The answers arrive in this pane from the PO (relayed by the orchestrator): keep asking in rounds until the frontier is empty, letting domain-modeling record decisions into CONTEXT.md and ADRs as they land.
  3. Hand off to /to-spec: it is reserved, you cannot invoke it. When the frontier is empty, announce ready for /to-spec and stop. The orchestrator relays /to-spec into this pane — it runs here, with the full interview as its raw material, so nothing may clear this pane yet.
  4. Stop. After /to-spec publishes the spec, announce the spec path in one line and stop. A later ruling: <issue path> invocation is a spec ruling on a triaged defect: answer from the spec file, append the ruling as a ## Comments entry on that issue, and stop.
Show full SKILL.md (677 more words)Show less
PO

You are the decision-maker standing in for the user. The orchestrator pastes pending questions (grilling rounds, breakdown approvals, triage rulings), each with a recommendation where one exists. You answer them; you never grill, spec, or code.

  1. Default to the recommendation. Adopt it unless it contradicts CONTEXT.md, an ADR, or the feature's published spec.
  2. Decide, don't expand. Answer exactly what was asked. Never invent new decisions, widen scope, or re-open settled ones.
  3. Escalate real conflicts. If a recommendation genuinely conflicts with the docs, or has consequences you cannot judge for the user (breaking changes, data loss, security, public API), answer ESCALATE: <question id> — <one line why>; the orchestrator takes that question to the user.

Output one line per question, no preamble: Q1: <decision> …

Dev

ticket: <NN> means implement that one ticket. One invocation, one deliverable. Files are your only interface: read .scratch/<feature-slug>/spec.md and issues/; never rely on conversation state from an earlier ticket. Ticket breakdown is not yours: the orchestrator sends /to-tickets into this pane directly, and the PO approves the breakdown here.

  1. Read ticket <NN> and its blockers. Set Status: in-progress and save before any work.
  2. Follow the /implement discipline with the ticket as your spec — /implement itself is reserved, but its primitives are yours: call the Skill tool for tdd where possible, at the ticket's pre-agreed seams; run typechecking regularly, single test files regularly, and the full test suite (pnpm test) once at the end; once done, call the Skill tool for code-review on the diff and address what it finds.
  3. Commit to the current branch, scoped to this ticket's files only; leave unrelated changes uncommitted.
  4. Set Status: done and append a ## Comments entry with the commit SHA.
  5. Announce ticket <NN> done (<sha>) and stop. Never start the next ticket — the orchestrator clears this pane and paces the next one.
QA

Your deliverables are .scratch/<feature-slug>/acceptance.md and defect tickets under issues/. You never fix code.

  1. Load context: read spec.md, every file in issues/, team.md (Base SHA = your review fixed point), and CONTEXT.md.
  2. Derive the checklist: one scenario per user story and per user-observable ticket acceptance criterion. Write acceptance.md from the template below with every scenario unchecked and Status: in-progress. On a regression run (focus: given) skip this — work from the existing file.
  3. Execute every scenario. Automated: pnpm test, pnpm build. CLI: pnpm skills-manager … against a disposable hub — a temp directory or the repo's my-skill-home/ — never the operator's real ~/.skills-manager. Manual-only: walk the documented flow and record what you observed. Record Result (pass / fail→F-id / blocked) and Evidence (the command plus its deciding output line, or the observed behavior). A pass without evidence is not a pass.
  4. Review the diff against the spec: call the Skill tool for code-review with fixed point = Base SHA and the spec path — both axes run; do not re-implement the Spec axis yourself. Turn its Spec findings into scenarios or defects; carry the Standards findings into the report only (dev already ran per-ticket review).
  5. File defects: for each fail, create issues/<NN>-<slug>.md (next free number) with **Category:** bug and Repro / Expected / Actual from your evidence. Confirmed defect → Status: ready-for-agent. Ambiguous (spec silent, not reproducible) → Status: needs-triage with a ## Triage Notes comment. Link the finding to its issue. Never check the box of a failed scenario.
  6. Close the run: all scenarios pass, no open defects → check every box, set Status: passed, append a one-paragraph summary (scenarios and commits covered, defects found and fixed, leftover risks). Otherwise set Status: failed and stop. After dev fixes or a spec amendment you are re-prompted with focus: finding ids: re-run those scenarios plus pnpm test, update evidence in place, then close the run again (this step).
<acceptance-template>
# Acceptance: \<feature-slug\>

Status: in-progress Spec: spec.md · Base: <sha> · Run: <n> · Date: <date>

Scenarios

  • S1: <user-observable scenario> Source: spec US-3 / ticket 02 AC-2 Result: pass | fail→F1 | blocked Evidence: \<command\> → <deciding output line>

Findings

F1: <title>

Scenario: S2 · Issue: issues/06-<slug>.md (ready-for-agent) Expected: … Actual: … Repro: … </acceptance-template>

Note: acceptance Status is QA's own three states — in-progress, failed, passed — not a triage label.

© shenysun, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file (references) in .claude/skills/team of shenysun/skills-manager.

  • SKILL.md
  • references/orchestration.md

Open the folder on GitHubat commit 3cf755f

Compare with similar skills

Team next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Team compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Team this skillshenysun/skills-manager197—~2.7kAutomated safety check: PassMIT
MCP Server Builderanthropics/skills180k63 repos~2.3kAutomated safety check: PassApache-2.0
Hook Development for Claude Code Pluginsanthropics/claude-plugins-official38k10 repos~4.1kAutomated safety check: NotesApache-2.0
Using Superpowersfarm-fe/farm5.6k36 repos~1.4kAutomated safety check: PassMIT
Executing Plans Inlineobra/superpowers297k2 repos~5.1kAutomated safety check: PassMIT
Skill CreatorAzure/azqr79689 repos~8.2kAutomated safety check: PassApache-2.0

Similar skills

  • MCP Server Builder

    anthropics/skills

    Official

    Guides the design and implementation of Model Context Protocol servers in TypeScript or Python, from tool naming and error messages to evaluation.

    180k GitHub starsUsed in 63 repos~2.3k tokens
    Agent WorkflowsAuto-check passed

  • Hook Development for Claude Code Plugins

    anthropics/claude-plugins-official

    Official

    Explains how to write Claude Code plugin hooks, both prompt-based checks and bash commands, for events such as PreToolUse, Stop and SessionStart.

    38k GitHub starsUsed in 10 repos~4.1k tokens
    Agent WorkflowsAuto-check: notes

  • Using Superpowers

    farm-fe/farm

    A skill your agent uses when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions

    5.6k GitHub starsUsed in 36 repos~1.4k tokens
    Agent WorkflowsAuto-check passed

  • Executing Plans Inline

    obra/superpowers

    Has the agent carry out an implementation plan itself, task by task in the current session, keeping a ledger, proving each step with a test and ending with one whole-branch review.

    297k GitHub starsUsed in 2 repos~5.1k tokens
    Agent WorkflowsAuto-check passed

  • Skill Creator

    Azure/azqr

    Official

    Create new skills, modify and improve existing skills, and measure skill performance.

    796 GitHub starsUsed in 89 repos~8.2k tokens
    Agent WorkflowsAuto-check passed

  • Claude Code Agent Development

    anthropics/claude-plugins-official

    Official

    Explains how to write agents for Claude Code plugins: the markdown file with YAML frontmatter, trigger descriptions, model and color settings, and system prompt design.

    38k GitHub starsUsed in 7 repos~2.8k tokens
    Agent WorkflowsAuto-check passed

More from shenysun/skills-manager

  • Skills Manager

    shenysun/skills-manager

    Operate the skills-manager CLI — manage a local skill hub (install, import, distribute to agents/projects, update) and backfill provenance for source-less skills (adopt lockfile evidence, search the…

    197 GitHub stars~7k tokensUpdated 23 days ago
    Auto-check passed

Categories

Questions about Team

What does Team do?

Run the four-role Herdr team (pm, po, dev, qa panes) that ships one feature of this repo. Team is an agent skill from shenysun/skills-manager. Run the four-role Herdr team (pm, po, dev, qa panes) that ships one feature of this repo.

When should I use Team?

Team fits situations like: agent Workflows work in your project.

How do I install Team in Claude Code?

Run `npx skills add shenysun/skills-manager --skill team -a claude-code`. Or copy the skill folder (.claude/skills/team in shenysun/skills-manager) into .claude/skills/team in your project. Claude Code loads it when a task matches its description.

How do I install Team in Codex?

Run `npx skills add shenysun/skills-manager --skill team -a codex`. Or copy the skill folder (.claude/skills/team in shenysun/skills-manager) into .agents/skills/team in your project. Codex loads it when a task matches its description.

Can I use Team in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add shenysun/skills-manager --skill team -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/team, .gemini/skills/team, .github/skills/team and .opencode/skills/team in your project.

What does Team need to run?

Going by SKILL.md and its folder, Team needs the command-line tools its instructions call (pnpm and git).

Does Team access the network?

SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Team safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Team use?

Team is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Team use?

About 2.7k tokens (SKILL.md is roughly 11k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 1.3k tokens, read only when the agent opens those files.

What are the alternatives to Team?

Skills that share tags, products or a category with Team: MCP Server Builder (anthropics/skills, 180k stars), Hook Development for Claude Code Plugins (anthropics/claude-plugins-official, 38k stars), Using Superpowers (farm-fe/farm, 5.6k stars) and Executing Plans Inline (obra/superpowers, 297k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Team?

shenysun (a GitHub user) maintains it in shenysun/skills-manager, which has 197 GitHub stars. The repository holds 2 skills in this directory. The repository was last updated on September 18, 2026.

Source: shenysun/skills-manager on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.