skill-state-manager

A Meta Skill that manages API keys and state persistence for other tools. Invoke whenever a tool requires authentication or fails with 401/403 errors.

This skill claims to simplify credential handling but leaves important security details out. Before installing or enabling it consider: (1) Ask the skill author to declare the exact config paths in metadata and explain how files are secured. (2) Do not paste highly sensitive secrets (AWS keys, private keys, service account JSONs) into chat unless you trust the skill and know how they will be stored. (3) Prefer secure stores (OS keyring, encrypted file, or platform-provided secret manager) over plaintext JSON; request that the skill set file permissions (600) or use encryption. (4) Require the skill to restrict prompting to only the minimal fields necessary and show the user exactly what will be written before writing. (5) If you cannot verify these protections, avoid using this skill for high-value credentials or disable autonomous invocation so prompts require explicit user action. Providing the skill with a clear design for secure storage (and updating metadata) would reduce the concerns raised here.

Type: OpenClaw Skill Name: skill-state-manager Version: 0.0.1 The 'skill-state-manager' skill (SKILL.md) acts as a meta-tool that instructs the AI to intercept authentication failures from other skills and prompt the user to provide sensitive credentials (API keys, tokens) directly within the chat interface. It then directs the AI to store these secrets in plaintext JSON files at predictable local paths, such as `~/.ai-skills-state/` or `~/.claude/skills-state/`. While presented as a convenience for managing tool state, this pattern introduces significant security risks by centralizing unencrypted secrets in well-known locations and encouraging insecure secret-handling practices.