The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data in the form of OpenAPI specifications, which could contain malicious instructions designed to influence agent behavior.
Ingestion points: The agent uses Glob, Grep, and Read tools to find and ingest local files matching patterns like **/openapi.{json,yaml,yml}. It also retrieves specifications from external Postman workspaces via MCP tools.
Boundary markers: The skill instructions do not specify the use of delimiters or provide instructions for the agent to ignore embedded prompts found within the specification's descriptions, summaries, or examples.
Capability inventory: The skill is granted access to the Bash tool (shell execution) and Write tool (file modification). A malicious specification could attempt to trick the agent into executing arbitrary commands or writing malicious files to the local environment.
Sanitization: There are no explicit instructions for the agent to sanitize or escape the content of the API specifications before analysis.
[COMMAND_EXECUTION]: The skill requires the Bash tool to search for files and perform its analysis workflow. While this is a standard requirement for the skill's stated purpose, the availability of a shell environment increases the potential impact of a successful indirect prompt injection attack.

postman-api-readiness