final-task-report
Use when the main execution agent must produce a final Markdown delivery report after the task is complete, especially when reviewer status is done and the…
Skill by waltstephen
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/waltstephen-final-task-report
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents, productivity, agent-building
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/final-task-report
What's inside
- Skill name
final-task-report- Description
- Use when the main execution agent must produce a final Markdown delivery report after the task is complete, especially when reviewer status is done and the report must be saved to a specific local file for human/mobile reading and bot delivery.
- License
- MIT
- Instructions
- 60 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Final Task Report Use this skill only in the final handoff phase. ## Goal Produce one Markdown report file that a user can read directly on phone or desktop. The report must stay grounded in the actual work completed in the repository and current session. Do not invent experiments, datasets, or code changes. This skill is meant to be updated over time as the reporting format evolves. Main agent should follow this skill directly for final report generation. ## Required Sections Always include these sections in order: 1. `## 0. Original User Task` or `## 0. 原始用户任务` 2. `## 1. What This Task Proposed` or `## 1. 本次实验提出了哪些点` 3. `## 2. How It Was Done and What Data Was Used` or `## 2. 具体是怎么做的,使用了什么数据` 4. `## 3. What the Main Agent Changed, What Approach It Used, and What Experiments It Ran` or `## 3. main agent 修改了什么,使用了什么思路,跑了什么实验` 5. `## 4. Special Notes` or `## 4. 特殊说明` If the task context is mixed-language, bilingual section titles are acceptable. ## Content Rules 1. Summarize only completed work. 2. If no dataset was used, write `未使用` or `none`. 3. If no experiment was run, write `未运行额外实验` or `none`. 4. Include the original user task and any explicit user follow-up requirements. 5. Special Notes should capture: - maintenance rules - user-visible caveats - follow-up constraints - any special thing the user explicitly asked to be highlighted 6. Do not infer or mention unrelated internal skill context. User requirements are enough. 7. Prefer the language of the user task: - Chinese task -> Chinese report - English task -> English report - mixed context -> bilingual or mixed headings 8. Prefer concrete evidence: - files changed - commands/checks run - observed outputs - validation results 9. Keep it readable on mobile: - short paragraphs - flat bullets - avoid long walls of text ## Write Contract 1. Write to the exact file path requested by the caller. 2. Do not rename the output file. 3. Do not create multiple alternative reports unless explicitly requested. 4. After writing, reply only with the final file path/status if the caller asks for a terse machine-readable response.
From waltstephen/ArgusBot, MIT.
More from waltstephen/ArgusBot
- 0.620score
planner-manager-explorer
waltstephen
Use when a planning or manager agent must maintain architecture, framework decomposition, TODO tables, and next-session objectives while also exploring…
SkillDFreshness: D★ 315Claude Code +9 - 0.620score
autonomous-research-loop
waltstephen
Operate Codex as a high-autonomy repository executor for implementation, debugging, and long-running research loops.
SkillDFreshness: D★ 315Claude Code +9 - 0.620score
pptx-run-report
waltstephen
Use this skill any time a .pptx file is involved — creating, reading, editing, or generating a run report presentation.
SkillDFreshness: D★ 315Claude Code +9 - 0.536score
btw-file-return
waltstephen
Detect when a BTW question is asking for project files, screenshots, or effect images, then return the most relevant local file paths quickly.
SkillDFreshness: D★ 315Claude Code +9
Signals
- Stars
- 315
- Forks
- 29
- Open issues
- 0
- Last commit (this folder)
- Mar 20, 2026 (7 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/waltstephen/ArgusBot
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from waltstephen
- 0.620score
planner-manager-explorer
waltstephen
Use when a planning or manager agent must maintain architecture, framework decomposition, TODO tables, and next-session objectives while also exploring…
SkillDFreshness: D★ 315Claude Code +9 - 0.620score
autonomous-research-loop
waltstephen
Operate Codex as a high-autonomy repository executor for implementation, debugging, and long-running research loops.
SkillDFreshness: D★ 315Claude Code +9 - 0.620score
pptx-run-report
waltstephen
Use this skill any time a .pptx file is involved — creating, reading, editing, or generating a run report presentation.
SkillDFreshness: D★ 315Claude Code +9 - 0.536score
btw-file-return
waltstephen
Detect when a BTW question is asking for project files, screenshots, or effect images, then return the most relevant local file paths quickly.
SkillDFreshness: D★ 315Claude Code +9
Cite this page
final-task-report skill. Agent Mart. https://agentmart.dev/skills/waltstephen-final-task-report (last updated 2026-10-24)