codingheader/myskills
Overview
This skill implements a Manus-style file-based planning workflow to manage complex, multi-step tasks. It automatically creates and maintains task_plan.md, findings.md, and progress.md so important context is persisted to disk. Use it to structure long-running work, coordinate research, and avoid lost context across many tool calls.
How this skill works
On start, the skill creates three markdown files and seeds them with templates. During execution it writes decisions, discoveries, and session logs to the appropriate files whenever key events occur. It enforces rules like creating the plan first, saving findings after every two information-gathering actions, and logging all errors to drive iterative fixes.
When to use it
- Starting a multi-step project with 3+ phases or many tool calls
- Conducting research that produces discoverable artifacts and notes
- Building or integrating systems where progress must be auditable
- Tasks expected to span multiple sessions or collaborators
- Any workflow where losing intermediate context would be costly
Best practices
- Always create task_plan.md before taking major actions
- Save discoveries to findings.md after every two view/search operations
- Read task_plan.md before making major decisions or changing direction
- Log every error, attempt, and result to progress.md and task_plan.md
- Follow the 3-strike protocol: diagnose, try alternatives, then rethink
Example use cases
- Researching and drafting a technical whitepaper across multiple sessions
- Coordinating a multi-step engineering task with build/test cycles and error handling
- Running a long investigative workflow that requires collecting sources and notes
- Developing a feature that requires iterative trials, logs, and decision checkpoints
- Managing a cross-team project where persistent artifacts support handoffs
FAQ
It creates task_plan.md for phases and decisions, findings.md for research and discoveries, and progress.md for session logs and test results to keep critical context persistent and searchable.
How often should I write to the files?
Write to findings.md after any discovery and at least once after every two view/browser/search actions. Update task_plan.md after each phase and log session activity continuously in progress.md.
14 skills
This skill helps you manage complex multi-step tasks by creating and updating task_plan.md, findings.md, and progress.md to persist project state.
This skill helps you debug LangChain and LangSmith agents by fetching recent traces and quickly analyzing execution patterns for insight.
This skill refactors inline styles and utilities into semantic CSS components with dark mode support and tests to improve maintainability.
This skill lets Claude perform real actions across apps like sending emails, creating issues, posting messages, and updating data.
This skill guides modern Next.js 15 frontend development with React 19, TS, Tailwind, and Server/Client patterns for scalable apps.
This skill helps you orchestrate complex reasoning tasks by coordinating specialized sub-agents to analyze, design, and critique multi-perspective solutions.
This skill analyzes images and diagrams to describe UI elements, extract text, note confidence, and flag issues.
This skill reviews existing CSS for semantic naming, Tailwind @apply usage, dark mode coverage, and tests, providing actionable feedback.
This skill guides iterative task planning before implementation, detailing objectives, approach, steps, verification, and clarifying questions to align with
This skill helps you create and manage Claude Code skills, hooks, agents, and commands, enabling streamlined infrastructure configuration.
This skill proactively searches memory for prior discussions, decisions, and context before starting new work, ensuring continuity and informed actions.
This skill analyzes tweets against Twitter's ranking signals and rewrites drafts to maximize reach and engagement.
This skill proactively searches auto-generated documentation to identify function signatures, classes, and APIs before implementing tasks.
This skill helps extract readable transcripts from Claude Code and Codex CLI sessions into markdown, enabling quick review and search.
