The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch external documentation from orpc.dev (e.g., https://orpc.dev/llms-full.txt) and explicitly tells the agent to trust the fetched pages over its own internal instructions or memory. This creates a vulnerability where instructions embedded in external content could override agent behavior during the migration process.
Ingestion points: Defined in SKILL.md, including https://orpc.dev/llms.txt, https://orpc.dev/llms-full.txt, and various documentation slugs.
Boundary markers: Absent. The instructions do not provide delimiters or warnings to ignore potentially embedded instructions in the documentation.
Capability inventory: The agent is expected to use standard file system operations for code migration and shell commands for dependency management.
Sanitization: Absent. No validation or filtering is specified for the content retrieved from the external URLs.
[COMMAND_EXECUTION]: The skill contains instructions to execute shell commands to manage the migration environment.
Evidence: Instructions to run npm install for various @orpc/* packages and npm view @orpc/server dist-tags to determine the latest version of the software.
[EXTERNAL_DOWNLOADS]: The skill references several external domains to obtain documentation and software packages.
Evidence: References to orpc.dev, v1.orpc.dev, and the NPM registry for package updates.

orpc-migrate