The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local binaries, including git and ruby, via subprocess.run to perform repository metadata lookups and strict YAML validation. These executions are performed using list-style arguments without a shell, which prevents command injection vulnerabilities.
Evidence: scripts/okf_yaml.py calls ruby with a hardcoded script; scripts/insert_frontmatter.py calls git log and git rev-parse.
[INDIRECT_PROMPT_INJECTION]: The skill processes external Markdown files which serve as a potential attack surface for indirect prompt injection.
Ingestion points: The skill reads all Markdown files identified as 'concepts' within the target directory specified in the root argument.
Boundary markers: The skill relies on standard YAML frontmatter delimiters (---) to separate metadata from document bodies.
Capability inventory: The skill has the capability to write to the local file system (using atomic replacement via os.replace) and execute local tools (git, ruby).
Sanitization: The skill performs extensive validation, including resolving paths to prevent directory traversal (mutation_path), checking for symlinks, and using a strict Ruby-based YAML parser (Psych) with safe_load to prevent malicious payloads in frontmatter.

x9-okf-docs