The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill's operations are confined to reading and writing local documentation files within a designated directory structure.
[COMMAND_EXECUTION]: The skill utilizes a specialized command /emit-pending-decisions to handle unresolved logic gaps. This tool is part of the intended workflow for managing human-in-the-loop decisions and prevents the AI from making autonomous assumptions.
[PROMPT_INJECTION]: The skill processes untrusted data from local markdown files, creating an indirect prompt injection surface.
Ingestion points: Reads decisions.md, seed.md, and other artifacts within docs/threads/ as described in SKILL.md.
Boundary markers: None identified for delimiting external file content.
Capability inventory: Performs file writes to spec.md as described in SKILL.md and executes the /emit-pending-decisions tool.
Sanitization: No explicit sanitization or validation of the ingested markdown content is performed.

reconcile-spec