The Agent Skills Directory

[COMMAND_EXECUTION]: The script assets/garden.py uses subprocess.run to call the local git binary. These calls are used to perform routine repository operations like checking the current HEAD SHA, verifying if the worktree is dirty, and generating diff statistics between versions. This is functional and limited to the tool's purpose of drift detection.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests content from OKF bundle files (explainer markdown files) to determine source status. This creates a surface where an attacker-controlled bundle could include malicious instructions for the agent within its metadata or descriptions.\n
Ingestion points: Reads YAML frontmatter and markdown content from explainer.md and index.md files.\n
Boundary markers: The skill does not implement specific delimiters for the content it relays to the agent.\n
Capability inventory: The tool performs file reads, git operations, and provides playbooks for writing file updates and executing site regeneration tools.\n
Sanitization: Uses a custom-built, strict parser for frontmatter that ignores non-essential YAML features.

knowledge-gardener