The Agent Skills Directory

[SAFE]: The bootstrap script scripts/init_local_vault.py includes a security-hardening step that explicitly clears core.hooksPath from Git configuration. This prevents the execution of unauthorized Git hooks that might be present in a cloned repository.
[COMMAND_EXECUTION]: The skill's Python scripts wrap system commands for git, gh, and obsidian. These commands are invoked using list-based arguments with subprocess.run and do not use shell=True, which effectively prevents shell injection vulnerabilities.
[SAFE]: The instructions in SKILL.md mandate explicit user confirmation before cloning any repository and enforce strict project-scoping guardrails. These measures ensure that the agent operates only on intended data and within authorized boundaries.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it reads and processes Markdown content from the Obsidian vault. Ingestion points: scripts/github_knowledge_skill.py (read command) and scripts/local_obsidian_knowledge.py (obsidian read). Boundary markers: None specified for vault content processing. Capability inventory: Subprocess execution of Git, GitHub CLI, and Obsidian. Sanitization: No explicit sanitization of Markdown content is performed before processing.

obsidian-gh-knowledge