The Agent Skills Directory

[SAFE]: The skill instructions define a legitimate workflow for course modification using a predefined set of tools such as list_scenes, read_stage, and patch_stage. No malicious command execution or unauthorized access patterns were identified.
[SAFE]: The skill does not contain hardcoded credentials, obfuscated code, or persistence mechanisms. All operations are performed through authenticated tool calls within the agent environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-generated course content which represents an ingestion surface for indirect prompt injection. The skill instructs the agent to read existing content before performing edits, which could expose it to malicious instructions embedded in the course data.
Ingestion points: Course page content retrieved via the read_stage tool in SKILL.md.
Boundary markers: None specifically defined to isolate untrusted course content from the agent instructions.
Capability inventory: The agent has capabilities to modify deck structure (edit_deck), patch slide elements (patch_stage), and regenerate audio (generate_tts) as described in SKILL.md.
Sanitization: The skill relies on structured JSON Pointer updates for most operations, which provides a layer of data validation.

pro-editing