Borrowing it
Nothing to install: this file belongs to JeremyDev87/codingbuddy. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/JeremyDev87/codingbuddy/master/.claude/skills/taskmaestro/SKILL.mdgit clone --depth 1 https://github.com/JeremyDev87/codingbuddyWrote 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/skills/jeremydev87/codingbuddy/taskmaestro)<a href="https://agentmods.dev/skills/jeremydev87/codingbuddy/taskmaestro"><img src="https://agentmods.dev/badge/skills/jeremydev87/codingbuddy/taskmaestro/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/skills/jeremydev87/codingbuddy/taskmaestro"><img src="https://agentmods.dev/badge/skills/jeremydev87/codingbuddy/taskmaestro.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 | $0.00028 | $0.11250 |
| Opus 5.5 | $0.00011 | $0.04500 |
| Sonnet 5.5 | $0.00006 | $0.02250 |
| Haiku 4.5 | $0.00003 | $0.01125 |
Grade D, and why
taskmaestro scanned grade D with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
cat ~/.claude/taskmaestro-state.json 2>/dev/null Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
rm -rf "$WT_PATH" How it starts
The opening of the file, as written. The whole thing — 1,012 lines — stays where its author put it; the contents beside it link to each section on GitHub.
TaskMaestro — tmux 패널 오케스트레이터
현재 tmux 세션의 기존 패널들에서 Claude Code를 병렬로 실행하고, 이 패널(지휘자)에서 제어한다. 각 패널은 독립된 git worktree에서 작업하며, 지휘자가 작업 지시, 상태 감시를 수행한다.
전제 조건: 지휘자(이 Claude Code)가 tmux 패널 중 하나에서 실행 중이어야 하며, 나머지 패널들이 이미 열려있어야 한다.
Arguments 파싱
$ARGUMENTS를 파싱하여 서브커맨드와 인수를 분리한다:
start [--repo <path>] [--base <branch>] [--panes <1,2,3>] [--review-pane] [--no-review-pane]assign <패널번호> "<작업 지시>"statuswatchsend <패널번호> "<텍스트>"stop [패널번호|all]
서브커맨드에 해당하는 섹션으로 이동한다.
현재 tmux 환경 감지
모든 서브커맨드에서 먼저 실행:
# 현재 tmux 소켓, 세션, 윈도우 정보를 자동 감지
TMUX_SOCKET=$(tmux display-message -p '#{socket_path}')
# 소켓 이름 추출 (경로에서 마지막 부분)
SOCKET_NAME=$(basename "$TMUX_SOCKET")
SESSION=$(tmux display-message -p '#{session_name}')
WIN_IDX=$(tmux display-message -p '#{window_index}')
MY_PANE=$(tmux display-message -p '#{pane_index}')
tmux 명령어에서
-L플래그는 소켓 이름을 사용한다.tmux -L "$SOCKET_NAME" send-keys ...형태로 사용.
Worker Completion Signaling (RESULT.json)
워커가 작업을 완료하면 .taskmaestro/wt-N/RESULT.json 파일을 생성하여 지휘자에게 결과를 알린다.
이 파일은 watch 모드의 1차 감지 수단이며, capture-pane은 fallback으로만 사용한다.
RESULT.json 스키마
{
"status": "success | failure | error | review_pending | review_addressed | approved",
"issue": "#767",
"pr_number": 123,
"pr_url": "https://github.com/org/repo/pull/123",
"timestamp": "2026-03-22T01:13:00.000Z",
"cost": "$0.45",
"error": null,
"review_cycle": 0,
"review_comments": []
}
| 필드 | 타입 | 필수 | 설명 |
|---|---|---|---|
status |
"success" | "failure" | "error" | "review_pending" | "review_addressed" | "approved" |
✅ | 작업 결과. success: PR 생성 완료 (리뷰 대기), failure: 작업 실패 (테스트 실패 등), error: 예기치 못한 오류, review_pending: 지휘자 리뷰 코멘트 작성 완료 (워커 응답 대기), review_addressed: 워커가 리뷰 반영 완료 (재리뷰 대기), approved: 지휘자 최종 승인 (진짜 완료) |
issue |
string | null |
❌ | 관련 이슈 번호 (예: "#767") |
pr_number |
number | null |
❌ | 생성된 PR 번호 |
pr_url |
string | null |
❌ | 생성된 PR URL |
timestamp |
string |
✅ | ISO 8601 완료 시각 |
cost |
string | null |
❌ | 세션 비용 (예: "$0.45") |
error |
string | null |
❌ | 에러 메시지 (status가 "error"일 때) |
review_cycle |
number |
❌ | 현재 리뷰 사이클 횟수 (기본값: 0) |
review_comments |
string[] |
❌ | 지휘자가 남긴 리뷰 코멘트 요약 목록 |
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 · 1,012 lines · 28 tokens per session scan D b7efac3677fe
taskmaestro is a skill published in the GitHub repository JeremyDev87/codingbuddy (31 stars, last pushed 6mo ago), licensed MIT. It adds 28 tokens to every session and 11,250 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it D with 2 findings (reads agent configuration directories, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
feature
Full per-feature pipeline: implement → testloop → evaluator challenge → coder feedback → document → commit.
ds-ship
Ship orchestrator — classify the project, pick a mode (improve/release/launch/maintain), delegate only the skills the project's signals justify, consolidate findings, produce an audit report. Use for an end-to-end audit or ship pipeline across multiple skills.
draft-issue
Draft a GitHub issue based on a discussion or requirements in the current conversation.
close-task
Close the task: update tracking and optionally its GitHub issue.
finish-task
Deprecated alias for close-task.
run-task
Run pending steps, then aftertask finalization; does not close the task.