The Agent Skills Directory

[PROMPT_INJECTION]: The skill describes workflows for ingesting and processing data from external providers and workspace files, creating an attack surface for indirect prompt injection.
Ingestion points: Data enters the context via provider-api-request, web-request, and workspaceRead from external APIs and the local file store as detailed in SKILL.md.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are mentioned in the provided examples.
Capability inventory: The workflows utilize high-risk tools including run-code (execution), workspaceWrite (file modification), and provider-api-request (network access) as listed in SKILL.md.
Sanitization: The skill does not provide guidance on sanitizing or validating external content before it is processed by the agent.
[COMMAND_EXECUTION]: Instructions are provided for utilizing the run-code tool for data processing tasks, including background execution for long-running scripts.
[EXTERNAL_DOWNLOADS]: The skill details the use of provider-api-request and web-request to fetch external data, supporting large payload transfers and automated pagination.

analysis-workspace