The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted data from multiple external sources and incorporates it into the document drafting process without sufficient safeguards.\n
Ingestion points: SKILL.md (Stage 1) specifies reading content from shared documents, fetching data from URLs, and ingesting information from communication platforms such as Slack and Microsoft Teams.\n
Boundary markers: Absent. The instructions do not require the agent to use delimiters or explicit "ignore embedded instructions" warnings when handling retrieved external content.\n
Capability inventory: The skill uses create_file and str_replace to perform file operations and can invoke sub-agents or search connected tools as described throughout SKILL.md.\n
Sanitization: No validation, escaping, or filtering is applied to the data gathered from external integrations before it is processed by the agent.

doc-coauthoring