The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill retrieves a service account token from the macOS login keychain using the security find-generic-password command. It documents a specific platform limitation where tokens passed via the -w flag during setup are briefly visible in the system process list.
[COMMAND_EXECUTION]: The skill invokes local command-line utilities including the 1Password CLI (op), the macOS security tool, and jq for structured data processing.
[DYNAMIC_EXECUTION]: The skill allows the execution of multi-line shell scripts using op run -- bash -s. It mitigates risks by requiring quoted heredocs (<<'SCRIPT') to ensure scripts are interpreted literally without parent shell expansion.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Reads item titles, vault names, and metadata from 1Password via op item list and op vault list, which are external data sources.
Boundary markers: The skill uses JSON formatting and jq for structural filtering, but lacks explicit delimiters or instructions to the agent to ignore instructions embedded within vault metadata.
Capability inventory: The skill has the ability to execute shell commands and write to the environment of child processes via op run.
Sanitization: Uses jq for output selection and basic text transformation (e.g., ascii_downcase), providing basic structure to the data ingested from external vaults.

1password