The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard shell commands for file management and Git operations. These include mkdir, printf, mv, tar, ln, git init, git add, git commit, git submodule, git clone, and git push. These operations are used to maintain the wiki structure, archive files, and synchronize with remote repositories.
[EXTERNAL_DOWNLOADS]: The skill facilitates the mounting of external wikis or repositories through git clone and git submodule add. It also suggests the installation of companion skills (obsidian-markdown and obsidian-bases) via npx skills based on user consent. These interactions target user-provided URLs or well-known ecosystem extensions.
[INDIRECT_PROMPT_INJECTION]: The skill processes project files, Git logs, and external data to extract knowledge, which presents a surface for indirect prompt injection.
Ingestion points: Project files and Git history are processed during "Project sync" mode. The scripts/ws.py script reads index.md files and page content to audit the wiki structure.
Boundary markers: The skill defines a strict trust scope distinguishing between "owned" and "external" folders (e.g., shared/). Instructions direct the agent to extract "durable knowledge" rather than re-synthesize entire conversations.
Capability inventory: The skill has capabilities for file system writes, Git network operations, and execution of a local Python helper script.
Sanitization: The ws.py script employs robust parsing logic using fenced_mask and frontmatter_end to ensure that markdown code blocks and metadata are not misinterpreted as navigation or structural commands. It also implements path normalization and reserved name checks to prevent directory traversal attacks.

ws-update