Personality Dynamics

Adapts OpenClaw agents by learning user interaction patterns, switching personas by context, and evolving personality weekly.

This skill appears to implement a local persona-evolution workflow (analyze local memory files, create PERSONA files, weekly reports) which is coherent with its description, but there are a few things to check before installing or running it: - Inspect and remove/adjust the hardcoded path in heartbeat-check.sh (cd /Users/nealme/clawd). That path is user-specific and could cause unintended behavior; do not run that script as-is. Replace with workspace-relative paths if you intend to use it. - Understand what local files it reads/writes: analyze-session.ts reads memory/<today>.md and weekly-report.ts suggests edits to SOUL.md. Ensure any files in your workspace that contain secrets or private data are not unintentionally exposed or processed. - The skill writes persistent data under PERSONA/ and PERSONA/evolves/. Review those outputs and their permissions; consider running the scripts in an isolated project directory first. - The CLI runs local scripts via execSync and the scripts print a spawn marker (---SPAWN_PERSONA_GENERATION---) intended for OpenClaw; review how your agent platform would act on that output before enabling autonomous invocation. - There is no declared homepage/source in the registry entry even though package.json references a GitHub URL; verify the package origin (confirm repository & publisher) before trusting automatic usage. - The interactive onboarding asks about permissions like drafting external communications, but the code does not itself perform network actions. If you plan to extend the skill to take external actions (send emails, post tweets), require explicit, auditable opt-in and add proper API credential handling. If you want to proceed: run the CLI scripts in a throwaway or sandbox workspace first, remove or rewrite heartbeat-check.sh, and verify that generated outputs and any 'auto-apply' behaviors require manual approval (the onboard config shows ask_before_major: true, which is good). If you are unsure, prefer forking and sanitizing the repo (remove hardcoded paths, add clear logging and permission prompts) before installing widely.

Type: OpenClaw Skill Name: personality-dynamics Version: 1.0.0 The skill is classified as suspicious due to significant vulnerabilities. The `cli.ts` script uses `execSync` to run other scripts with unsanitized user arguments (`remainingArgs`), posing a shell injection risk (RCE vulnerability). Additionally, `generate-persona.ts` directly embeds user input into LLM prompts that are explicitly spawned by the OpenClaw agent, creating a prompt injection vulnerability against the agent's underlying LLM. Other scripts like `expand.ts` and `onboard.ts` also write unsanitized user input into persona definition files, which could serve as indirect prompt injection vectors if the agent's LLM reads these files for context. These are critical vulnerabilities, not evidence of intentional malicious behavior.