The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions and code involve running a local Python script (scripts/run.py) to process workspace files. This script performs file reading and writing operations to generate the table_schema.md artifact.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local workspace files, which acts as a potential surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: scripts/run.py reads GOAL.md, outline/outline.yml, outline/subsection_briefs.jsonl, and outline/evidence_drafts.jsonl to populate the generated schema.
Boundary markers: Absent. The script interpolates content from GOAL.md directly into the generated markdown without specific boundary markers or instructions for the agent to ignore embedded content.
Capability inventory: The script performs file writing (atomic_write_text) to the local workspace.
Sanitization: Absent. The script performs basic presence checks for the 'goal' string but does not sanitize or escape the content for safe interpolation into downstream prompts.

table-schema