The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes course metadata, including titles, descriptions, and HTML overviews, from the vendor's API. This data is externally controlled and could potentially contain instructions aimed at influencing the agent's behavior.
Ingestion points: Untrusted data enters the context via GET /api/ibl/v1/course_settings and GET /settings/advanced/{COURSE} in SKILL.md.
Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the course content it reads.
Capability inventory: The skill can perform network operations using curl and execute local scripts.
Sanitization: There is no evidence of validation or escaping for the data interpolated from course settings.
[DYNAMIC_EXECUTION]: The skill uses Python one-liners to perform data manipulation tasks like URL encoding and JSON parsing at runtime.
Evidence: SKILL.md contains instructions to use python3 -c for encoding course keys and extracting asset URLs from API responses. These are considered low risk as they use standard library functions from known templates.
[COMMAND_EXECUTION]: The skill relies on executing local scripts and system tools to manage authentication and network requests.
Evidence: The skill calls node .claude/skills/iblai-api-studio-auth/scripts/studio-login.mjs --check for session verification and uses curl to interact with the ibl.ai infrastructure.

iblai-api-studio-settings