The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute administrative commands, including doppler run --command patterns which evaluate shell strings containing injected environment variables.
[COMMAND_EXECUTION]: The AWS_WORKFLOW.md file includes instructions for destructive operations, specifically aws iam delete-access-key, which can lead to service disruption if executed improperly.
[DATA_EXFILTRATION]: The SKILL.md file contains patterns for reading sensitive files from the local filesystem, specifically ~/.claude/.secrets/gh-token-accountname, which constitutes data exposure of credentials.
[DYNAMIC_EXECUTION]: Employs mise [env] configuration that uses cache(run='...') to dynamically execute shell commands (e.g., doppler secrets get) to populate environment variables at runtime.
[EXTERNAL_DOWNLOADS]: References the installation of the Doppler CLI via Homebrew (brew install dopplerhq/cli/doppler). This targets a well-known service and is documented neutrally.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it ingests untrusted data from Doppler secrets and the filesystem without explicit sanitization before interpolating them into shell commands.
Ingestion points: Doppler secrets values/notes, .mise.toml, and files in ~/.claude/.secrets/.
Boundary markers: Absent; commands are executed directly in the shell context.
Capability inventory: High-privilege actions including aws iam delete-access-key, doppler secrets set, and arbitrary Bash execution.
Sanitization: No evidence of validation or escaping for external content before execution.

doppler-workflows