The Agent Skills Directory

EXTERNAL_DOWNLOADS (HIGH): The installation instructions in SKILL.md use curl -sLf ... https://get.doppler.com | sh. Piped execution of remote scripts from untrusted domains is a high-risk pattern that allows unverified code to run directly on the user's system.
REMOTE_CODE_EXECUTION (HIGH): The skill allows the agent to run arbitrary commands using doppler run -- <command>. Because this command injects sensitive workplace secrets into the process environment, it can be exploited to execute malicious payloads with elevated access.
CREDENTIALS_UNSAFE (MEDIUM): The skill is designed to retrieve and handle sensitive secrets. Commands such as doppler secrets get and doppler secrets download can expose API keys, database URLs, and other credentials within the agent's context and logs.
COMMAND_EXECUTION (MEDIUM): The skill uses the Bash tool to perform sensitive operations on the Doppler platform, including project deletion and service token management. Misuse of these tools could lead to significant infrastructure compromise.
PROMPT_INJECTION (LOW): The skill has an indirect prompt injection surface.
Ingestion points: Secret values and notes retrieved from the Doppler API via doppler secrets get (SKILL.md).
Boundary markers: None; untrusted data from secrets is not delimited or marked to be ignored.
Capability inventory: Bash command execution, WebFetch network access, and Write file access.
Sanitization: None; the skill does not provide mechanisms to sanitize or escape data retrieved from secrets.

doppler