The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies extensively on shell command execution and heredocs to automate credential management tasks.
AWS_WORKFLOW.md and references/aws-credentials.md contain multiple blocks of bash scripts using /usr/bin/env bash << 'EOF' to interact with the AWS CLI and Doppler CLI.
The scripts perform sensitive operations such as creating access keys, piping them to Doppler, and generating audit reports.
[INDIRECT_PROMPT_INJECTION]: The skill defines high-privilege workflows that ingest untrusted user input which is then interpolated into shell commands, creating a vulnerability surface.
Ingestion points: Placeholders for <username>, <profile>, and <SECRET_NAME> in AWS_WORKFLOW.md and references/aws-credentials.md.
Boundary markers: Absent. The skill instructions do not include specific delimiters or warnings to the agent to sanitize these inputs before interpolation.
Capability inventory: The skill utilizes aws (IAM, S3, STS), doppler (secrets, activity), jq, and file system tools (shred, rm).
Sanitization: Absent. There is no logic provided to escape or validate user-provided strings before they are executed in a shell context.

doppler-workflows