The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill instructions direct the agent to read and load environment variables from a sensitive file at an absolute local path: /Users/danielhowells/.codex/plugins/secrets/1password-service-account.env. This file is expected to contain a OP_SERVICE_ACCOUNT_TOKEN used for credential management.
[EXTERNAL_DOWNLOADS]: The skill relies on an external command-line utility, starlingcli, hosted in a repository belonging to the skill author (github.com/howells/starlingcli).
[COMMAND_EXECUTION]: The skill executes shell commands using op run and starlingcli to retrieve bank balances and transactions. It includes logic to bootstrap the environment by sourcing a local credential file and setting environment variables for Starling Bank tokens.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes data from an external source (Starling Bank API via starlingcli).
Ingestion points: Output from starlingcli commands such as balance and transactions (SKILL.md).
Boundary markers: None identified; the agent is instructed to use the structured JSON output directly.
Capability inventory: The agent can execute subprocesses via op run and starlingcli (SKILL.md).
Sanitization: No explicit sanitization or filtering of the bank's JSON responses is specified before the agent processes them.

starling