The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting data from external sources and using it to inform the execution of design jobs.
Ingestion points: Untrusted data is retrieved from external sources like CSV files, Google Sheets, or Notion databases as specified in the sources field of job files defined in SKILL.md.
Boundary markers: The instructions do not specify any delimiters or boundary markers to prevent the agent from obeying instructions embedded within the ingested data.
Capability inventory: The skill utilizes start_open_design_job_run and finish_open_design_job_run to execute a "brief" which acts as a set of natural language instructions for the design process.
Sanitization: There is no evidence of content sanitization or validation for the data pulled from external connectors before it is processed by the agent.
[EXTERNAL_DOWNLOADS]: The documentation within the skill directs users to download and execute a setup utility from a remote repository.
Evidence: The skill recommends running npx @feimacode/open-design-agent-kit init --tool claude to initialize the environment.
Context: The package is a vendor-owned resource originating from the skill's author (feimacode).
[DYNAMIC_EXECUTION]: The skill dynamically generates a job file (.open-design/jobs/<id>.job.md) that contains natural language instructions (the "brief"), which are interpreted and executed at runtime during the "Run" phase.
Evidence: The "Set up" section describes writing the job file and subsequently executing it via start_open_design_job_run.

open-design-recurring