The Agent Skills Directory

[PROMPT_INJECTION]: Metadata poisoning detected in SKILL.md. The author is listed as 'OpenAI', which is a deceptive claim as the actual skill author is 'tristanmanchester'. This misrepresentation leverages the reputation of a trusted organization to gain undue trust in the skill's safety.
[PROMPT_INJECTION]: Vulnerability to indirect prompt injection through data processed from the Todoist API.
Ingestion points: The skill retrieves user-controlled data such as task content, descriptions, and comments via scripts/todoist_api.py (e.g., in do_get_tasks and do_get_comments).
Boundary markers: No delimiters or instructions are used to distinguish untrusted data retrieved from the API from system prompts.
Capability inventory: The skill possesses significant capabilities, including the ability to create, update, and delete tasks and projects, and to execute arbitrary API requests via the raw command in scripts/todoist_api.py.
Sanitization: There is no evidence of sanitization or filtering of retrieved content in scripts/todoist_api.py before it is incorporated into the agent's context.

todoist-api