The Agent Skills Directory

[COMMAND_EXECUTION]: The skill describes workflows involving Git operations and the execution of code on remote hosts. These are standard developer tasks intended to support persistence and remote testing.
[DATA_EXFILTRATION]: The skill includes explicit security guidelines to prevent the accidental exposure of sensitive data, such as .env files and credentials, during the synchronization process.
[PROMPT_INJECTION]: The skill processes project source code and chat transcripts across environments. The documentation provides clear instructions to avoid including sensitive values in these transcripts, which serves as a mitigation against data leakage during cross-environment processing.
[INGESTION_POINTS]: Project source code and chat transcripts from the local environment.
[BOUNDARY_MARKERS]: Mentions the use of explicit omission markers for binary attachments and warnings regarding unsafe context.
[CAPABILITY_INVENTORY]: Git operations (push/fetch) and code execution on a paired remote host.
[SANITIZATION]: Explicit 'Don't' rules against copying .env files or tokens into the metadata or relay.

portal