Planning With Files

File-based planning for complex tasks. Use persistent markdown files as working memory to survive context resets. Creates task_plan.md, findings.md, and progress.md. Use for any task requiring >5 tool calls, research projects, or multi-step implementations.

This skill is internally coherent and appears to do what it says — create and maintain three markdown files as persistent planning memory. Before installing or running it, consider: 1) The planning files (task_plan.md, findings.md, progress.md) will be created in your project root and persist on disk (and in your git history if committed). Do not log secrets or sensitive data there. 2) Review the included scripts (init-session.sh, check-complete.sh) — they are simple shell scripts that create/check files, but you should inspect them and run in a safe folder first. 3) If you use version control, consider adding these files to .gitignore or keeping them in a sandbox workspace to avoid accidental leakage. 4) The package source/homepage is unknown; although there are no network endpoints or remote installs in the package itself, prefer installing from a trusted source or inspecting the files locally before adding to long-running agents. 5) The agent can invoke skills autonomously by default; if you plan to run this skill in an autonomous agent, be mindful that it will create/modify files without prompting. If you want further checks (e.g., to ensure it never writes outside a specific directory), provide that constraint and I can point out the exact lines to change.

Type: OpenClaw Skill Name: planning-files Version: 1.0.0 The OpenClaw AgentSkills bundle 'planning-files' is designed to enhance an AI agent's planning and memory by utilizing local markdown files as persistent storage. All scripts (`scripts/check-complete.sh`, `scripts/init-session.sh`) perform benign local file operations (creating, reading, and parsing markdown files) strictly within the stated purpose. The `SKILL.md` and `references/manus-principles.md` contain instructions and guidelines for the AI agent that promote disciplined workflow, error handling, and context management, without any evidence of prompt injection designed for malicious activities like data exfiltration, unauthorized execution, or ignoring user instructions. The `README.md` includes installation instructions, one of which (`npx add https://github.com/wpank/ai/tree/main/skills/meta/planning-with-files`) is non-standard but does not indicate malicious intent within the skill's operational logic. Overall, the skill aligns with its stated purpose and lacks any high-risk behaviors or malicious indicators.