The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local project files, creating a potential surface for indirect instructions.
Ingestion points: Reads the Outline file (markdown), AGENTS.md (config), and supervisor/MEMORY.md (status/issues).
Boundary markers: Relies on a regex (Result pattern) for data extraction, but lacks explicit "ignore embedded instructions" warnings for the markdown content processing.
Capability inventory: File system read access and file write access for report generation.
Sanitization: The skill focuses on extracting specific fields (figure number, panels, messages) rather than executing content, which naturally mitigates most risks.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use shell commands for directory management.
Evidence: Uses mkdir -p to ensure the Report output dir exists before writing the summary file. This is a standard, low-risk administrative operation within the project scope.

todofig