Ogp Project

Tool-agnostic project collaboration for AI assistants. Users keep their own tools (Linear, Jira, Obsidian, GitHub, iCloud, local files — anything). This skil...

This skill is coherent with its stated purpose but exercises networked collaboration and local state access — before installing: (1) Inspect and, if needed, backup the local files it will read (~/.ogp*, ~/.ogp-hermes*). (2) Review the npm package @dp-pcs/ogp and its GitHub repo (the SKILL.md points to github.com/dp-pcs/ogp) to ensure you trust the publisher and install scripts. (3) Be aware the skill's mandatory 'pre-task' checks will cause your agent to query peer agents automatically when project-related triggers fire — if that could leak sensitive work-in-progress, restrict triggers or disable autonomous invocation. (4) Confirm the list of approved peers in your OGP config so the agent only contacts trusted endpoints.

Type: OpenClaw Skill Name: ogp-project Version: 0.6.0 The skill facilitates multi-agent project collaboration by using a custom CLI tool (ogp) to share project metadata and user intent with 'peer agents'. It is classified as suspicious because it requires broad access to local configuration files (e.g., ~/.ogp/projects.json, ~/.ogp-hermes/config.json) and mandates 'Proactive Pre-Task Checks' that automatically transmit user activity and project context to remote peer IDs. While these behaviors support the stated goal of cross-agent synchronization, the automated exfiltration of sensitive project details (repository paths, notes locations, and task lists) to external entities via the @dp-pcs/ogp package presents a significant data exposure risk.