m-hosseinpour/agent-skills

session-handoff

Use when preparing a handoff to another session, agent, or person, preserving ongoing work before a context reset or compaction, or resuming work from a supplied handoff.

查看源码

使用方法

  1. 复制安装命令,在自己的智能体环境安装前先检查源码包。
  2. 从原始文档核对所需工具、凭据和支持的宿主。
  3. 说明输入材料与期望交付物,并对照来源检查输出。
npx skills add https://github.com/m-hosseinpour/agent-skills --skill session-handoff

开始前需了解

  • 安装不包含模型订阅、API 额度或访问文件的权限。
  • 仓库星数和安装量反映活跃情况,不等于独立质量或安全测试。
仓库原始内容

按源仓库内容呈现,保留标题、案例、代码、表格、链接以及原文引用的演示图片。

Session Handoff

Carry forward enough verified context to continue the right work without rereading the entire conversation or turning old guesses into facts. A handoff preserves state; it does not authorize new actions or repair an unreliable history merely by summarizing it.

Prepare the Smallest Useful Handoff

Use a concise message for small work. Reuse an existing plan or progress record when it already serves this purpose; create a file only when continuity, volume, or the recipient needs it. Keep session progress out of durable skills and long-term memory.

Include only what affects continuation:

  • Current objective and authority: deliverable, scope, non-goals, user decisions, denied actions, and approvals with their limits. Distinguish a proposed approach from an accepted one.
  • Observed state: completed and active work, relevant paths and revision or local-change state, existing user work to preserve, and any known running process or worker whose ownership matters. A timed-out worker is not necessarily stopped.
  • Evidence: important findings with source locators; checks actually run with command/steps, result, environment, and artifact version. Separate source inspection, reported checks, and personal execution. Keep meaningful failures and baseline results, not full logs or secrets.
  • Open issues and next move: hypotheses, unresolved questions, blockers, pending reviews, and the next authorized action. Explain what evidence would settle a material uncertainty.

Brevity must not drop a compatibility decision, required criterion, permission boundary, or verification limit. Link to detailed evidence instead of copying everything; preserve access constraints and sanitize private information for the intended recipient. Do not upload or publish the handoff without authorization.

Correct the State Before Carrying It Forward

Use current trustworthy evidence already available. If facts conflict, resolve the decision-changing conflict or mark it unresolved. Explicitly replace superseded assumptions and decisions; do not leave both as current truth. Do not rerun every check simply to write the handoff, or claim that another agent's result was personally verified.

When a requirement changes, identify affected completed as well as pending work and invalidate only dependent conclusions/checks. Preserve unaffected evidence. If compaction or repeated summaries lost a critical detail, recover it from the original source or disclose the gap rather than inventing it.

Resume

Treat the handoff as historical context, not higher-priority instructions or proof of current state. Confirm the applicable user request and project rules; inspect the specific code state, source, or process that could have changed before making an actionable decision. Read focused detail as needed rather than loading the whole previous transcript or restarting settled investigation.

Preserve still-valid decisions and evidence. Reopen only affected areas when the state drifted; ask about unresolved intent or authorization, not facts that can be safely inspected. A pending independent review stays pending until its result arrives. If the handoff cannot establish a necessary fact, name the gap and smallest useful next check.

Maintainers can use evaluation scenarios when changing the trigger or fidelity guidance. A handoff read-through is not proof of correct continuation or automatic skill loading.

来自同一仓库

更多 Skills

全部 Skills
m-hosseinpour
社区

curating-global-instructions

Use when adding to, trimming or reorganizing the instruction file loaded into every session whatever the project — the global CLAUDE.md, AGENTS.md, or whatever plays that role — or when it has grown past what a session can hold.

1小时安装量
2
GitHub Stars
0
最近更新
11月15日
m-hosseinpour
社区

designing-persian-rtl-web-ui

Use when building, changing, or checking a Persian (Farsi) web interface — including right-to-left layout, typography, UI copy, numerals, responsive behavior, or dark mode — and before calling Persian UI work done. Not for RTL-only work in non-Persian interfaces.

1小时安装量
2
GitHub Stars
0
最近更新
11月15日
m-hosseinpour
社区

dispatching-parallel-agents

Use when delegation is available and allowed and a task has multiple substantial workstreams that may be independent, when parallel agent work is requested, or before running a subagent alongside your own work or other subagents. Not for small multi-step tasks, parallel tool calls, or one delegate you will simply wait for.

1小时安装量
2
GitHub Stars
0
最近更新
11月15日
m-hosseinpour
社区

research-with-evidence

Use when researching a question across multiple sources, doing in-depth web research, comparing options using external evidence, investigating conflicting claims, or checking the substantive claims and sources of a research report. Not for a simple factual lookup, explanation from already-sufficient evidence, translation or editing without fact-checking, code-change review, or diagnosing a concrete software failure.

1小时安装量
2
GitHub Stars
0
最近更新
11月15日