Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add TeamManagementPlugin/claude-plugin/plugin install team-managementWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/teammanagementplugin/claude-plugin/init)<a href="https://agentmods.dev/commands/teammanagementplugin/claude-plugin/init"><img src="https://agentmods.dev/badge/commands/teammanagementplugin/claude-plugin/init/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/teammanagementplugin/claude-plugin/init"><img src="https://agentmods.dev/badge/commands/teammanagementplugin/claude-plugin/init.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00019 | $0.03088 |
| Opus 5.5 | $0.00008 | $0.01235 |
| Sonnet 5.5 | $0.00004 | $0.00618 |
| Haiku 4.5 | $0.00002 | $0.00309 |
Grade B, and why
init scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads MCP configurationmediumAgent snooping
mcp.json carries server URLs and auth tokens; reading it lets a mod discover and abuse other integrations.
`.mcp.json`) that invoke the literal token **`python3`** — correct on macOS/Linux. How it starts
The opening of the file, as written. The whole thing — 191 lines — stays where its author put it; the contents beside it link to each section on GitHub.
team-management: Init
Set up team-management for THIS project: enable the plugin on this machine, wire the
behavioral guidance into the project's CLAUDE.md via native @-includes (durable
across /compact, unlike one-shot SessionStart injection), and optionally enable the
LLM Wiki.
Two different sharing models are in play, by design:
- Plugin enablement is per-machine.
.claude/is gitignored wholesale by the SessionStart hook (ensure_claude_dir_gitignored), so.claude/settings.jsonis NOT committed and does NOT auto-enable the plugin for teammates — each developer runs/team-management:initthemselves and puts their own tokens in the per-project.claude/state/provider-tokens.jsonfile. - The behavioral-guidance files ARE committed.
CLAUDE.tm.md,team-management/knowledge/*.md, and the managed@-block inCLAUDE.mdlive at the project root / underteam-management/(outside the.claude/ignore), so the whole team gets the same guidance via the committedCLAUDE.md@-includes.
Workflow
0. Windows runtime bootstrap (Windows only — provisions python3)
The plugin's hooks and MCP server launch from static manifests (hooks.json /
.mcp.json) that invoke the literal token python3 — correct on macOS/Linux.
Windows has no python3.exe (only python.exe / the py launcher), so on Windows
those launches fail (python3: command not found) and the plugin is dead until
python3 resolves to a real executable. A python3.cmd/.bat shim is NOT enough:
Claude Code spawns the MCP command directly, and Windows cannot direct-spawn a .cmd
without a shell — only a real .exe is found by every launch path (direct-spawn, bash,
cmd, PowerShell). Do this FIRST — before the config_update step below, which needs
the runtime up.
- Detect the platform. On macOS/Linux skip this whole section. Treat as Windows
when
unamereportsMINGW*/MSYS*/CYGWIN*, or%OS%isWindows_NT, or only PowerShell/cmd is available. - Is
python3already a real, runnable.exe? Runpython3 --versionANDwhere python3(where.exe python3). Skip the rest ONLY if--versionprints 3.x and the firstwhere python3hit ends in.exe. If it instead resolves to a.cmd/.batshim (broken for the MCP server per above, even though--versionworks), opens the Microsoft Store, or errors → do NOT skip; continue and provision a real.exebelow. - Require the
pylauncher. Runpy --version. Ifpyis missing → STOP and tell the user to install Python from python.org with the py launcher option (or the Microsoft Store build, which already provides a realpython3.exe), then re-run/team-management:init. Do not substitute another interpreter. - Find the Python install dir (holds
python.exe, is on PATH, and — for the python.org installer — precedes%LOCALAPPDATA%\Microsoft\WindowsApps, so the shim is not shadowed by a Storepython3.exeapp-execution-alias):py -3 -c "import os,sys;print(os.path.dirname(sys.executable))". - Create a REAL
python3.exeby copyingpython.exein that same directory. A copy ofpython.exeIS a workingpython3— it loads the samepythonXX.dllfrom its own directory — and is directly spawnable by the MCP launcher. Create it only ifpython3.exeis absent there. Use whichever form works in the agent's shell (the detection commands above are plain program calls and work in any shell; only this copy is shell-specific):- PowerShell:
Copy-Item "<dir>\python.exe" "<dir>\python3.exe" - cmd:
copy /Y "<dir>\python.exe" "<dir>\python3.exe"(the/Ykeeps it non-interactive — plaincopypromptsOverwrite?and would stall the agent shell) - Git-bash:
cp "<dir>/python.exe" "<dir>/python3.exe"
- PowerShell:
- Verify in a fresh process:
python3 --versionmust print a 3.x version andwhere python3must point at thepython3.exeyou just created in the dir from step 4 (NOT a WindowsApps alias). Ifwhere python3finds nothing, that dir is not on PATH (Python was installed without "Add Python to PATH"), so copying the exe there cannot help — tell the user to either re-run the python.org installer with "Add Python to PATH" enabled, OR add the dir to PATH (setx PATH "%PATH%;<dir>", then open a NEW shell), and then re-run/team-management:init. Otherwise print the exact manual copy command and stop. - Tell the user to RESTART Claude Code, then continue setup. On a fresh Windows
install the hooks (including the
config_intent_gate.pyhook that authorizesconfig_update) AND the MCP server were already launched broken this session and only retry on restart — so the wiki/guidance step below CANNOT complete now. After the restart, run/team-management:configto finish it.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 191 lines · 19 tokens per session scan B 1956d86f83a7
init is a command published in the GitHub repository TeamManagementPlugin/claude-plugin (6 stars, last pushed 2d ago), licensed MIT. It adds 19 tokens to every session and 3,088 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it B with 1 finding (reads mcp configuration). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
save
Save current work state for next session. Creates/updates .planning/ files (CHECKPOINT.md, STATE.md, settings.json) so Heimdall resumes with full context. NOT a rewind — saves forward progress. Run before closing a session or at any milestone.
status
Show current Heimdall state — project phase, active agents, quality gates, conflicts.
loops
Weekly review — surface everything in-flight and force a ship / block / kill decision on each.
p4-sprint
Plans the current sprint: pulls matching stories from the backlog, defines the sprint goal, identifies risks, and creates the sprint document. Called repeatedly at the start of each new sprint.
p4-backlog
Breaks down the features from FEATURES.md into actionable tasks, builds the prioritized backlog, and defines milestones and release planning. The result is BACKLOG.md and PROJECTPLAN.md as the steering foundation for the entire implementation.
p5-polish
Collects small clean-up TODOs accumulated during the sprint (from /p5-review, /p5-acceptance, /p5-bugfix, TODO/FIXME comments) that do not justify a full story but make the next sprint easier. Triages each item into polish-now, backlog, handover, or drop. Optionally executes polish-now items directly (TDD mini-cycle…