The Agent Skills Directory

[SAFE]: The skill is primarily instructional and does not request or use any sensitive tools or permissions.- [COMMAND_EXECUTION]: The skill includes local evaluation scripts in the evals/fixtures/ directory (check-domain-decisions.sh and check_domain_decisions.py). These are used for validating skill outputs during development and do not perform network operations or access sensitive system files.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided quality dashboard metrics and sketches (Ingestion point: user prompts). While there are no specific boundary markers mentioned for user input and no formal sanitization beyond data quality auditing, the skill lacks access to any tools (e.g., shell, network, file system) that would allow an injection to be exploited for harm beyond the immediate chat context (Capability inventory: generative text only; Sanitization: input audit for data quality).

quality-dashboard-design