The Agent Skills Directory

[SAFE]: The skill consists of educational reference files that promote defensive programming and secure networking practices. It provides patterns for robust API communication, such as implementing exponential backoff for retries and ensuring that logout logic correctly clears all derived session state.
[COMMAND_EXECUTION]: The skill includes a utility in references/curl-logger-ktor-plugin.md for generating curl commands from outgoing HTTP requests. It uses a correct POSIX shell quoting implementation (shellQuote) to ensure that replaying these commands is safe and does not lead to command injection if the request content contains shell-sensitive characters like single quotes or dollar signs.
[CREDENTIALS_UNSAFE]: The skill explicitly addresses credential safety. It provides guidance on redacting sensitive headers from debug logs and outlines a secure flow for handling OAuth callbacks, ensuring that tokens are delivered directly to session storage rather than through potentially insecure navigation parameters.
[INDIRECT_PROMPT_INJECTION]: While the skill establishes patterns for ingesting data from remote APIs (untrusted external sources), it emphasizes defensive parsing, normalization, and structural classification to maintain data integrity. The skill provides the necessary architectural boundaries to mitigate risks associated with processing external content.

remote-api-footguns