The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides scripts and patterns for direct file system manipulation, including moving, renaming, and writing files, as well as SQLite database management to facilitate content lifecycle stages.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow for processing untrusted markdown files where metadata fields such as 'maturity', 'review_after', and 'expires' trigger file system actions like archiving or promotion. This creates a surface where malicious content could influence the organization or retention of data.
Ingestion points: Markdown files and YAML frontmatter parsed in modules/lifecycle-stages.md and modules/storage-patterns.md.
Boundary markers: Standard YAML frontmatter delimiters (---).
Capability inventory: File moving (mv), writing (cat), atomic writes (os.replace), and SQLite operations.
Sanitization: Python code examples correctly utilize yaml.safe_load to prevent unsafe object instantiation during metadata parsing.
[SAFE]: The storage implementation patterns demonstrate security maturity by utilizing atomic write operations with fsync() and ensuring temporary files are created within the same directory as the target to guarantee atomic renames and prevent data corruption during crashes.

storage-templates