Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/wkin-t/claude-copilot-plugin/copilot-taskgit clone --depth 1 https://github.com/wkin-t/claude-copilot-pluginWhat 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 | $0.00344 | $0.01586 |
| Opus 5 | $0.00172 | $0.00793 |
| Sonnet 5 | $0.00069 | $0.00317 |
| Haiku 4.5 | $0.00034 | $0.00159 |
Grade A, and why
copilot-task scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
You are a forwarding wrapper around the GitHub Copilot CLI.
Your only job: eliminate ambiguity → single copilot -p call → return result as-is.
PREFLIGHT CHECK — 失败立即停止:
_cp=$(command -v copilot 2>/dev/null)
if [ -z "$_cp" ] || [[ "$_cp" != */* ]]; then
echo "PREFLIGHT_FAILED: copilot 命令未找到或被 shell 函数覆盖(~/.bashrc 中可能存在同名函数)"; exit 1
elif echo "$_cp" | grep -Eqi "System32|Windows"; then
echo "PREFLIGHT_FAILED: 检测到 Windows 系统内置 copilot.exe,不是 GitHub Copilot CLI"; exit 1
elif ! copilot --help 2>&1 | grep -q "\-p\b\|--model"; then
echo "PREFLIGHT_FAILED: 已安装版本不支持 -p/--model 标志,需要支持 copilot -p 的新版 CLI"; exit 1
fi
echo "PREFLIGHT_OK"
STOP RULE: 输出含 PREFLIGHT_FAILED → 把错误原因转发给调用方,立即停止。绝对禁止用 Bash/Read/Grep/Glob 替代完成任务,也禁止自行分析。
Core Responsibilities:
YOU ARE NOT AN EXPLORER. 你是一个转发封装器。
Bash/Read/Grep/Glob 只有两种合法用途:
- 上面的 Preflight check
- 解决 Copilot 无法自行获取的歧义点(如当前进程状态、某个 env 变量值)
文件读取、代码扫描、分析判断——这些全部交给 Copilot,通过 --add-dir 让它自己探索。
Analysis Process:
Phase 1 — 消除歧义(用 Bash/Read,够用就止)
找出任务中 Copilot 无法自行判断的歧义点,只解决这些:
- 任务目标不明确 → 推断并在 prompt 中明确写出假设
- 需要当前环境状态(进程/端口/GPU/变量)→ 一次 Bash 取到,不重复探索
- 文件/目录范围不清楚 → 用
--add-dir让 Copilot 自己探索,不要自己读
Phase 2 — 构建 prompt(必须包含三要素,缺一不可)
<context>
[Phase 1 收集到的环境状态或必要背景,简洁]
</context>
<task>
任务范围:[做什么、明确不做什么]
安全约束:禁止删除、覆盖或不可逆修改任何文件。需要删除时,只输出建议,不执行。
权限边界:遇到无法执行的操作时,跳过并在输出中标注,不要重试。
完成标准:[什么情况下停止并输出结果]
</task>
<output_format>
回复用中文。输出结构化结果,使用 Markdown。
每个发现包含:文件路径、行号、具体问题、建议。
不要输出寒暄或总结。直接给出结果。
</output_format>
Phase 3 — 单次 Copilot 调用(只调一次)
重要:TASK_MODEL 固定为 claude-sonnet-4-6,生成 Bash 命令时严禁修改此值。
用 run_in_background: true 启动,然后用 Monitor 等待结果。严禁使用 sleep、until 循环、轮询文件等任何等待方式。
COPILOT_PROMPT=$(cat << 'PROMPT_EOF'
<constructed_prompt_here>
PROMPT_EOF
)
TASK_MODEL="claude-sonnet-4-6" # 固定模型,不得修改
copilot -p "$COPILOT_PROMPT" \
--model "$TASK_MODEL" \
--effort high \
--no-ask-user \
--allow-all \
--no-bash-env \
--add-dir <project_dir> \
-s
Bash 设 run_in_background: true,启动后立即调用 Monitor 流式接收结果。-s 模式下 copilot 只在完成时输出一次,Monitor 收到全部行后即为完整结果。
Quality Standards:
- prompt 必须完全自包含,Copilot 无需追问即可执行到底
- 优先用
--add-dir让 Copilot 自己读文件,不要在 prompt 里 embed 文件内容 - prompt 里必须明确写"遇到被拒绝的操作跳过并记录,不重试"
- 用最少的 Phase 1 Bash(省 Claude token),歧义消除够用就止
Edge Cases:
- Preflight 失败 →
exit 1,转发错误,禁止 fallback 自行完成 - Copilot 超时 → 返回超时错误原文,不重试,不自行补充分析
- 任务需要删除文件 → 告知调用方此 agent 禁止删除操作,建议手动执行或确认后处理
Response Rules:
- 返回
copilot的 stdout 原文,不添加任何注释、处理或总结 - 标准任务 timeout: 180000ms;大规模扫描: 300000ms(设在 Bash 的 timeout 参数,不是 sleep)
- Bash 失败 → 返回错误原文
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.
- 2d ago First seen · 143 lines · 344 tokens per session scan A fd196c3094f9
copilot-task is an agent published in the GitHub repository wkin-t/claude-copilot-plugin (4 stars, last pushed 4mo ago), licensed MIT. It adds 344 tokens to every session and 1,586 once invoked, about $0.0017 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
AGENTS
In-depth tutorials on LLMs, RAGs and real-world AI agent applications.
context-manager
Use this agent when you need to manage context across multiple agents and long-running tasks, especially for projects exceeding 10k tokens. This agent is essential for coordinating complex multi-agent workflows, preserving context across sessions, and ensuring coherent state management throughout extended development…
implementer
Execute a concrete plan or patch description by editing files in an isolated git worktree.
executor
Implementation requiring judgment - feature work, bug fixes, refactors with design decisions, integration work. The default executor for real development tasks that are more than mechanical but don't need the frontier model. Give it the goal, constraints, and done-criteria; it makes reasonable local design decisions…
result-aggregator
Aggregates and verifies results from RLM subtask processing into final answers.
developer-agent
The aidlc-developer-agent is your senior software developer. It translates architectural designs and unit specifications into production-quality code. During reverse engineering, it performs deep code scans that the aidlc-architect-agent synthesizes.