The Agent Skills Directory

[SAFE]: The skill performs legitimate project documentation tasks, primarily reading from the local specs/ directory and writing a summary file. No malicious behavior such as obfuscation, credential theft, or unauthorized network access was detected.
[COMMAND_EXECUTION]: The skill uses git log to determine directory creation dates. This is a standard project management function and is appropriately scoped to the relevant project paths to avoid unintended access.
[PROMPT_INJECTION]: The skill processes untrusted content from specification documents, creating a surface for indirect prompt injection.
Ingestion points: Reads various markdown files (e.g., requirements.md, tasks.md, design.md) within the specs/ subdirectories to extract summaries and check task status.
Boundary markers: None provided in the instructions; the agent is expected to parse the file content directly without specific delimiters or 'ignore' instructions.
Capability inventory: The skill can perform local file writes to specs/OVERVIEW.md and execute git commands to fetch metadata.
Sanitization: There are no instructions for escaping, validating, or sanitizing the content extracted from the files before it is interpolated into the summary document.

specs-overview