The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a significant surface for indirect prompt injection. It is designed to ingest and process external, untrusted content including workflow YAML files, facets (Markdown), and execution logs.
Ingestion points: Workflow YAMLs and facets are loaded in 'Step 1: Identify and Load Targets'.
Boundary markers: There are no explicit boundary markers or instructions to the agent to ignore potentially malicious embedded instructions in the processed files.
Capability inventory: The skill uses tools like Write, Edit, and Bash in various optimization and validation steps (e.g., implement and audit steps).
Sanitization: No sanitization or validation of the ingested content is mentioned before it is interpolated into agent prompts.
[COMMAND_EXECUTION]: The file scripts/validate-takt-files.sh contains a relative path reference (../../takt-piece-builder/scripts/validate-takt-files.sh) pointing to a location outside the skill's own directory. This constitutes an unverifiable external dependency that may execute unknown code if the referenced directory is present on the user's system.

takt-optimizer